Improved solidification scheme design method and system for roadbed soil body

Abstract

The invention provides an improved solidification scheme design method and system for a roadbed soil body, and relates to the field of soil body solidification, and the method comprises the steps: obtaining soil body solidification data, carrying out the expansion of the soil body solidification data through deep learning, and obtaining a solidification agent information database; screening in a curing agent information database according to the soil parameters to obtain a basic formula and corresponding strength data; constructing an initial relation equation based on the basic formula; coding level design of influence factors is carried out through central composite design, corresponding true values are obtained, and fitting is carried out on the initial relation equation to obtain an early strength equation and a late strength equation; and solving an optimal intersection of the solutions of the early strength equation and the late strength equation to obtain a target formula, and obtaining an improved curing scheme according to the target formula. According to the method, the optimal ratio of the raw materials is automatically calculated and output through deep learning and central composite design, so that intelligent soil curing agent formula design is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of soil solidification, and more specifically, to a method and system for designing an improved solidification solution for subgrade soil. Background Art

[0002] Soil solidifying agents are materials used to improve the properties of soil, which improve the engineering properties of soil through chemical reactions or physical actions when mixed with soil. These soil solidifying agents can improve the properties of soil by increasing the compressive strength and bearing capacity of soil, improving soil stability, reducing the liquidity index, and enhancing impermeability. At present, most soil solidification constructions select appropriate soil solidifying agents based on different types of base soils in existing implementation cases, but cannot predict or simulate the optimal proportioning schemes of soil solidifying agents that may exist outside the existing implementation cases. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for designing an improved solidification solution for subgrade soil to solve the above problems. To achieve the above purpose, the technical solutions adopted by the present invention are as follows:

[0004] A method for designing an improved solidification solution for subgrade soil includes:

[0005] Obtaining soil solidification data, where each piece of the soil solidification data includes soil parameters, the used solidifying agent formula, and soil compressive strength; expanding the soil solidification data through deep learning to obtain a solidifying agent information database;

[0006] Obtaining the soil parameters of the construction area, and screening in the solidifying agent information database according to the soil parameters to obtain a basic formula and corresponding strength data, where the basic formula includes the raw materials in the formula and their corresponding contents, and the strength data includes test strength data and predicted strength data;

[0007] Based on the basic formula, taking the soil solidification strength as the response value and the contents of each raw material as influencing factors, constructing an initial relationship equation;

[0008] Based on the basic formula, performing coded level design of the influencing factors through central composite design and obtaining corresponding true values, where the true values represent the contents of each raw material, and looking up the corresponding strength data according to the true values;

[0009] Fitting the initial relationship equation according to the true values and strength data and performing variance analysis, and performing elimination processing on the terms in the initial relationship equation based on the variance analysis results to obtain an early strength equation and a late strength equation;

[0010] Taking the maximization of the solidification strength of the soil mass as the goal, find the intersection of the solutions of the early strength equation and the late strength equation to obtain the target formula, and obtain the improved solidification plan according to the target formula.

[0011] On the other hand, the present application also provides a system for designing an improved solidification plan for subgrade soil masses, including:

[0012] A first acquisition module, configured to acquire soil solidification data, each piece of the soil solidification data including soil parameters, the curing agent formula used, and the soil compressive strength; expand the soil solidification data through deep learning to obtain a curing agent information database;

[0013] A second acquisition module, configured to acquire the soil parameters of the construction area, and screen in the curing agent information database according to the soil parameters to obtain a basic formula and the corresponding strength data, the basic formula including the raw materials in the formula and the corresponding contents, and the strength data including test strength data and predicted strength data;

[0014] A construction module, configured to construct an initial relationship equation based on the basic formula, with the soil solidification strength as the response value and the contents of each raw material as the influencing factors;

[0015] A design module, configured to perform coded level design of the influencing factors through central composite design based on the basic formula and obtain the corresponding true values, the true values representing the contents of each raw material, and find the corresponding strength data according to the true values;

[0016] A fitting module, configured to fit the initial relationship equation according to the true values and the strength data and perform variance analysis, and perform elimination processing on the terms in the initial relationship equation based on the variance analysis results to obtain an early strength equation and a late strength equation;

[0017] A first calculation module, configured to take the maximization of the soil solidification strength as the goal, find the intersection of the solutions of the early strength equation and the late strength equation to obtain the target formula, and obtain the improved solidification plan according to the target formula.

[0018] The beneficial effects of the present invention are as follows:

[0019] The present invention uses a deep learning model to expand the existing soil solidification data, enrich the data volume, and optimize and select the most suitable soil formula according to the soil conditions. In order to further determine the optimal ratio of each raw material in the soil formula, the present invention performs an optimization experiment through central composite design with multiple factors and multiple levels, automatically calculates and outputs the optimal ratio of the raw materials, thereby realizing intelligent design of the soil curing agent formula.

[0020] Other features and advantages of the present invention will be described in the subsequent specification, and in part will be obvious from the specification, or can be understood by implementing the embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments. It should be understood that the following drawings only show some embodiments of the present invention, and therefore should not be regarded as a limitation of the scope. For those of ordinary skill in the art, without creative efforts, other relevant drawings can also be obtained based on these drawings.

[0022] Figure 1 It is a flowchart of the design method for the improvement and solidification scheme of the subgrade soil body in the embodiment of the present application;

[0023] Figure 2 It is a schematic diagram of the three-dimensional calculation model of the foundation in the embodiment of the present application;

[0024] Figure 3 It is a schematic diagram of the layout of monitoring points in the embodiment of the present application;

[0025] Figure 4 It is a schematic diagram of the method model for the shallow solidification double-layer foundation in the embodiment of the present application;

[0026] Figure 5 It is a schematic diagram of the method model for the shallow solidification + plain concrete pile composite foundation in the embodiment of the present application;

[0027] Figure 6 It is a schematic diagram of the method for the shallow solidification + cement mixing pile composite foundation in the embodiment of the present application;

[0028] Figure 7 It is a schematic diagram of the system structure for the design of the improvement and solidification scheme of the subgrade soil body in the embodiment of the present application.

[0029] Markings in the figure: 100 - First acquisition module; 110 - Classification unit; 120 - First construction module; 130 - Expansion unit; 140 - Second construction unit; 200 - Second acquisition module; 210 - First acquisition unit; 220 - First screening unit; 230 - Second screening unit; 300 - Construction module; 400 - Design module; 500 - Fitting module; 600 - First calculation module; 700 - Prediction module; 800 - Second calculation module; 1000 - Generation module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0031] It should be noted that similar reference numerals and letters denote similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0032] See Figure 1 , this application provides a method for designing an improved solidification solution for subgrade soil, including steps S100, S200, S300, S400, S500, and S500;

[0033] Step S100: Obtain soil solidification data. Each piece of the soil solidification data includes soil parameters, the curing agent formula used, and the soil compressive strength; expand the soil solidification data through deep learning to obtain a curing agent information database;

[0034] Specifically, the soil solidification data can use the open-source relational database management system MySQL as the data storage platform. A series of data including complete soil parameters (soil density ρ, water content w, liquid limit W L , plastic limit W P ), the curing agent formula used (the raw materials of the curing agent used and the content of each raw material), and the soil compressive strength (including: 7-day unconfined compressive strength Y 7 and 28-day unconfined compressive strength Y 28 ) is used as 1 piece of the recorded soil solidification data. The source of each piece of data can be: ① The results of completed laboratory tests; ② The test results published in journals; ③ The test results of published patents.

[0035] The step of expanding the soil solidification data through deep learning to obtain a curing agent information database includes the steps:

[0036] S110: Traverse the soil solidification data, classify the data with the same soil parameters and the same curing agent raw materials, and obtain a soil solidification data packet for each type of soil;

[0037] S120. Perform deep learning based on each of the soil solidification data packets, using soil parameters and solidifying agent formulations as inputs and soil compressive strength as the output to construct a first prediction model;

[0038] The specific steps of deep learning are as follows: classify the data information in each "soil solidification data packet". Specifically, take the complete soil parameters and the used solidifying agent formulation as the input labels of the deep learning neural network. At the same time, take the corresponding soil compressive strength as the output label. After the input labels and output labels are both classified, randomly divide each "soil solidification data packet" into a training set, a validation set, and a test set according to the ratio of 7:2:1 in terms of the number of items.

[0039] The activation functions considered in the constructed deep learning neural network model can optionally include Sigmoid, ReLU, Tanh, and Leaky ReLU; the deep learning neural network model uses dynamic adjustment of the learning rate to improve the optimization effect, which can optionally include learning rate decay and cosine annealing; the optimizers that the deep learning neural network model can choose from include: stochastic gradient descent, Adam, Adagrad, and Adadelta. Moreover, to improve the construction efficiency of the deep learning neural network model, a deep learning neural network model construction program based on AI learning can be adopted. The deep learning neural network model construction program based on AI learning learns at least 1000 groups of deep learning neural network construction processes with a prediction accuracy of more than 90%. The AI learning content includes the input labels, the number of output labels, the determination process of the neural network model, the adjustment process of the number of neurons in the hidden layer, the adjustment process of the number of hidden layers, the adjustment process of the weight type, the determination process of the activation function, the adjustment process of the optimization algorithm model, etc. in the process of constructing a deep learning neural network with a prediction accuracy of more than 90%.

[0040] In this embodiment, a first prediction model is constructed for each soil solidification data packet, that is, each type of soil (with the same soil parameters) has its own exclusive first prediction model.

[0041] Train the deep learning neural network model constructed for each "soil solidification data packet", and use the test set to evaluate the performance of the trained model. It is required that the prediction accuracy of the deep learning neural network model corresponding to each "soil solidification data packet" is greater than 90%. At the same time, in the case where the prediction accuracy of the deep learning neural network model corresponding to the "soil solidification data packet" is lower than 90%, first record the architecture of the deep learning neural network model (including: neural network model, number of neurons in the hidden layer, number of hidden layers, weight type, activation function type, optimization algorithm model, etc.), reconstruct the deep learning neural network model corresponding to the individual "soil solidification data packet", and shield the recorded architecture of the deep learning neural network model until the prediction accuracy of the deep learning neural network model corresponding to the constructed "soil solidification data packet" is greater than 90%.

[0042] S130. Multiply the soil parameters and the curing agent formula by different preset coefficients to obtain extended parameters, and perform permutations and combinations on the extended parameters in sequence to obtain an extended parameter group. Input the extended parameter group into the first prediction model to obtain the predicted soil compressive strength:

[0043] As an example, identify the soil parameters and the curing agent formula used, and independently amplify each data by coefficients of 50%, 60%, 70%, 80%, 90%, 100%, 110%, 120%, 130%, 140%, and 150% to obtain extended parameters. At the same time, perform different permutations and combinations on each obtained extended parameter, list all permutation and combination cases, and obtain an extended parameter group (for example: 50%×ρ, 70%×w, 100%×W L 、120%×W P 、70%×P.O 42.5 grade cement content, 130%×fly ash content, 50%×basalt fiber content, 90%×polyvinyl alcohol resin content);

[0044] Input each extended parameter group into the first prediction model respectively to obtain the soil compressive strength under each parameter permutation and combination.

[0045] S140. Store all extended parameter groups and the corresponding predicted soil compressive strength to construct a curing agent information database.

[0046] S200. Obtain the soil parameters of the construction area, and screen in the curing agent information database according to the soil parameters to obtain the basic formula and the corresponding strength data. The basic formula includes the raw materials in the formula and the corresponding contents, and the strength data includes test strength data and predicted strength data; specifically includes the steps:

[0047] S210. Obtain the material information within the preset range of the construction area and determine the priority raw materials;

[0048] The situation of construction sites, factories and mines within 5 kilometers of the construction area can be captured in real time through UAV aerial photography technology and online map tools, so as to obtain information on available raw materials nearby and use them as priority raw materials. Considering factors such as resource conservation, environmental protection, and reduction of project construction costs, natural soil and construction waste (mainly inorganic non-metallic building solid waste that can be obtained nearby) can be selected as the raw materials for the curing agent.

[0049] S220. Screen in the curing agent information database based on the soil parameters to obtain the preliminary screened curing agent formula data;

[0050] Obtain the soil parameters input by the user, match them in the curing agent information database, and select the data in the database where the difference between each value of the soil parameters and the input soil parameters is not greater than 90% to obtain the preliminary screened curing agent formula data;

[0051] S230. Screen out the curing agent formula data including priority raw materials from the preliminary screened curing agent formula data, and then screen the curing agent formula data according to the corresponding soil compressive strength to obtain the basic formula and the corresponding strength data.

[0052] After screening out the curing agent formula data including priority raw materials, sort them in descending order according to the soil compressive strength corresponding to each data, and select the top 20% of the data; further, if the soil compressive strength of the selected data all meets the requirements, the information on the market price and transportation cost of each curing agent raw material obtained by calling a web crawler can be used to calculate the cost corresponding to each screened data, and select the data with the lowest cost to obtain the basic formula and the corresponding strength data. The screening results of this embodiment are: P.O42.5 grade cement: 2% - 8%; fly ash: 2% - 6%; basalt fiber: 0.1% - 0.5%; polyvinyl alcohol resin (PVA): 0.1% - 0.5% (mass percentage).

[0053] S300. Based on the basic formula, construct an initial relationship equation with the soil curing strength as the response value and the content of each raw material as the influencing factor;

[0054] Establish the initial relationship equation between the influencing factor and the response value as shown in Equation (1):

[0055]

[0056] In the formula, Y is the equation prediction value (set the 7-day unconfined compressive strength Y 7 and the 28-day unconfined compressive strength Y 28 as two groups of prediction values), k is the number of experimental influencing factors, β 0is the equation intercept, β i is the linear coefficient, β ij is the interaction coefficient, β ii is the squared coefficient, where i and j represent two different influencing factors, i = 1, 2, 3…k, j = 1, 2, 3…k.

[0057] S400. Based on the basic formula, perform the coded level design of the influencing factors through central composite design and obtain the corresponding true values, where the true values represent the contents of the raw materials, and find the corresponding strength data according to the true values; including the steps:

[0058] Determine the number of influencing factors for central composite design according to the number of types of raw materials in the basic formula; there are as many influencing factors as there are raw materials;

[0059] Determine the central value, high value, and low value of each influencing factor according to the content ranges of the raw materials in the basic formula; the high value and low value are the highest and lowest contents of the raw material, and the central value is the intermediate value of the raw material content range;

[0060] Determine the step size x of the level change of each influencing factor according to the preset coded level and the high value and low value of the influencing factor b :

[0061]

[0062] where x 低 is the low value, x 高 is the high value; the number of steps b 步数 is determined by the coded level, with a one-step interval between every two adjacent coded levels;

[0063] Calculate the true value corresponding to the coded level according to the coded level, step size of level change, and central value of each influencing factor.

[0064] The logical design of the central composite design optimization experiment with multiple factors and multiple levels includes the cube points, axial points, and central points of the full factorial or fractional factorial design;

[0065] The cube points (corner points) are the basis of the design and are usually a 2 k full factorial design, where k is the number of influencing factors. In more complex cases, a fractional design may be used to reduce the number of experiments required. In this embodiment, the influencing factor is the number of raw materials in the formula, and based on the basic formula screened in S230, k = 4;

[0066] The axial points (star points) are located on each factor axis at a certain distance from the design center and are usually represented by α. This distance can be fixed or optimized according to the needs of the experiment.

[0067] To meet the requirement of rotation, the value of α should be calculated according to Equation (2):

[0068]

[0069] The center point is at the central level of all factors. The main function of the center point is to estimate the experimental error and detect the curvature of the curve. Usually, the experiment is repeated multiple times to improve the accuracy of the estimation. The number of center points is recorded as C 0 . The number of center points is taken according to Table 1:

[0070] Table 1 Relationship between the number of influencing factors and the number of center points

[0071]

[0072] As can be seen from the above, the total number of designed experiments is calculated according to Equation (3):

[0073] Total number of experiments = 2 k + 2k + C 0 (3)

[0074] X i is the coded value parameter (i = 1, 2, 3... k), and each curing agent component (influencing factor) corresponds to an X i , in this embodiment, the curing agent raw materials determined in step S230 are: P.O42.5 grade cement (X 1 ): fly ash (X 2 ): basalt fiber (X 3 ): polyvinyl alcohol resin (X 4 ).

[0075] The specific value of X during design i is determined according to Equation (4):

[0076] X i = (x i - x 0 ) / Δx i (4)

[0077] In the formula, x i represents the true value of each influencing factor, x 0 represents the central value of each influencing factor, and Δx i represents the step size of the change in the level of each influencing factor.

[0078] In this embodiment, P.O42.5 cement, fly ash, basalt fiber, and polyvinyl alcohol resin (PVA) are preliminarily determined as the curing agent materials in the highway project. To obtain the optimal proportioning scheme under the interaction of the above four materials, an optimization experiment with 4 factors and 5 levels is designed. Substituting into formula (2) for calculation, the number of center points is taken as α = 2. Therefore, for the 4 influencing factors, P.O42.5 cement (X 1 ), fly ash (X 2 ), basalt fiber (X 3 ), and polyvinyl alcohol resin (X 4 ), encoding is performed according to formula (4), and the encoding levels are shown in Table 2. A test model is constructed, and the names, units, low values, and high values of each influencing factor are input, and C 0 = 6 is set. At the same time, 2 groups of predicted results (Y 7 and Y 28 ) are set. Specifically, taking P.O42.5 cement as an example:

[0079] The true value range of P.O42.5 cement is x 低 = 2% to x 高 = 8%; x 低 is the lowest content, and x 高 is the highest content;

[0080] Calculate the center value x 0 as:

[0081]

[0082] Calculate the step size of the level change x b :

[0083] The encoding levels -2, -1, 0, 1, 2 are given, so it is obtained that the step from -2 to -1 is 1 step, the step from -1 to 0 is 1 step, and so on. In the case of 5 encoding levels, the number of steps b 步数 = 4;

[0084]

[0085] The encoding levels -2, -1, 0, 1, 2 are given. According to formula (4), the true values corresponding to the encoding levels are calculated by back-calculation:

[0086] When X 1 = -2: x 1 = 5% + (-2)·1.5% = 2%;

[0087] When X 1 = -1: x 1 = 5% + (-1)·1.5% = 3.5%;

[0088] When X 1=0: x 1 =5%+0·1.5%=5%;

[0089] When X 1 =1: x 1 =5%+1·1.5%=6.5%;

[0090] When X 1 =2: x 1 =5%+2·1.5%=8%;

[0091] Table 2 Experimental influencing factors and coding levels

[0092]

[0093] The coding levels of each influencing factor are arranged and combined, and the soil compressive strength corresponding to different combinations is obtained. The results are shown in Table 3:

[0094] Table 3 Experimental plan and results

[0095]

[0096]

[0097] The coding levels of schemes No. 25-30 are all 0, but the corresponding curing strengths are different. This is because the curing agent information database contains data from multiple real parallel tests under the same ratio (in line with probability theory to avoid contingency), that is, test strength data, and also includes the data predicted by the first prediction model based on the ratio, that is, predicted strength data.

[0098] S500, fitting the initial relationship equation according to the true value and the strength data and performing variance analysis, eliminating the items in the initial relationship equation based on the variance analysis result, and obtaining an early strength equation and a late strength equation;

[0099] Based on the test scheme and results in Table 3, existing analysis tools can be used to automatically derive the relationship equation and perform variance analysis. The variance analysis results of the second-order model of the solidified soil for 7 days and the variance analysis results of the second-order model of the solidified soil for 28 days in this embodiment are shown in Tables 4 and 5, where the P value is an indicator used to evaluate the statistical significance of the differences between the data obtained from the experiment. It means whether the differences between different treatment levels are caused by random changes or by real influencing factors. Specifically, the P value represents the probability of a more extreme situation than the sample result when the null hypothesis is true. Therefore, a significance level d needs to be given in the test of models and factors. When P < d, the corresponding items are considered to have significant differences, and when P is greater than d, the corresponding items are considered to have no significant differences.

[0100] The F value is a statistic, F = between-group variance / within-group variance, which is used to test whether there are significant differences in the means between different groups.

[0101] Table 4 ANOVA results of the 7-day second-order model for solidified soil

[0102]

[0103] As can be seen from Table 4, the P value of the 7-day strength model for solidified soil is < 0.0001 < b, and R 2 = 0.9873, indicating a high model fitting degree; R 2 refers to the coefficient of determination, which is the square of the correlation coefficient. The coefficient of determination represents the proportion of the variance of the dependent variable that can be explained by the independent variable.

[0104] Analyzing the test results of each influencing factor again, it is found that X 1 X 3 、X 2 X 3 、X 2 X 4 、X 3 X 4 The corresponding P values are 0.7588, 0.6751, 0.7665, and 0.9734, all > b = 0.1. Therefore, it is considered that the effects of these four items on the 7-day strength of solidified soil are not significant, and they are removed. Finally, the 7-day strength equation of solidified soil, that is, the early strength equation, is given as shown in the formula:

[0105]

[0106] Table 5 ANOVA results of the 28-day second-order model for solidified soil

[0107]

[0108] As can be seen from Table 5, the P value of the 28-day strength model for solidified soil is < 0.0001 < b, and R 2 = 0.9524, indicating a high model fitting degree; Analyzing the test results of each influencing factor again, it is found that X 3 、X 1 X 3 、X 2 X 3 、X 2 X 4 、X 3 X 4The corresponding P-values are 0.5159, 0.3745, 0.6973, 0.6263, and 0.9050 respectively, all of which are > b = 0.1. Therefore, it is considered that the effects of these four items on the 7-day strength of the solidified soil are not significant, and they are removed. Finally, the 28-day strength equation of the solidified soil, that is, the late strength equation, is given as shown in the formula:

[0109]

[0110] Taking the maximization of the soil solidification strength as the goal, the intersection of the solutions of the early strength equation and the late strength equation is obtained to get the target formula, and the improved solidification plan is obtained according to the target formula. In this embodiment, the calculated target formula of the soil solidifying agent is: P.O42.5 grade cement: fly ash: basalt fiber: polyvinyl alcohol resin = 8.00: 4.91: 0.34: 0.46.

[0111] The optimal solidifying agent formula for a specific soil is determined through the above steps. In addition, the regulation of soil deformation needs to be considered to determine the specific solidification construction method. Therefore, the present invention also includes step S700: determining the soil solidification construction process based on numerical simulation, specifically as follows:

[0112] S710. Based on the target formula, the solidification performance information is obtained through a pre-established second prediction model. The solidification performance information includes the early solidification strength, early elastic modulus, late solidification strength, and late elastic modulus of the soil under different confining pressure conditions;

[0113] The confining pressure conditions can be 0 kPa (unconfined), 50 kPa, 100 kPa, 200 kPa, 300 kPa, 400 kPa, and 500 kPa;

[0114] S720. The cohesion parameter and internal friction angle parameter of the soil after solidification are obtained based on the solidification performance information; the cohesion parameter and internal friction angle parameter can be calculated through the Mohr-Coulomb strength criterion.

[0115] In this embodiment, the model parameters in Table 6 are calculated:

[0116] Table 6 Model Parameters

[0117]

[0118] S730. Based on the solidification performance information, cohesion parameter, and internal friction angle parameter, a three-dimensional model of the foundation in the construction area is constructed, and the settlement displacement of the foundation under different solidification construction methods is obtained through numerical simulation. The target solidification construction method is determined based on the settlement displacement;

[0119] The different solidification construction methods generally include: the shallow solidification double-layer foundation method, the shallow solidification + plain concrete pile composite foundation method, and the shallow solidification + cement mixing pile composite foundation method.

[0120] S740. Obtain an improved solidification plan according to the target formula and the target solidification construction method.

[0121] Obtain the settlement displacements under three solidification construction methods and compare them with the national codes, such as the "Code for Design of Highway Subgrade", as shown in Table 7.

[0122] Table 7 Allowable Post-construction Settlement of Soft Soil Subgrade (unit: cm)

[0123]

[0124] Select the solidification construction process with the lowest cost and meeting the design requirements.

[0125] The following is an illustration with specific examples: According to the on-site survey report, a certain section of land in a certain area was originally a lotus root planting field with weeds distributed in it. The surface was covered with 3 - 5 m thick silty clay, and the underlying layer was sandy mudstone intercalated with sandstone. The groundwater level in this section was relatively high, the soil was in a soft - firm plastic state, and it was a highly compressible soil. The maximum filling height of the embankment was about 10 m, and uneven settlement was likely to occur. The soil was improved and solidified, and the average treatment depth was 3.0 m. The overall width of the subgrade of the expressway project in this area was 34.5 m, the separated subgrade width was 2×17.0 m, the embankment filling height was 10 m, a 2 - m - wide platform was set at the middle height of 5 m to ensure the stability of the embankment. The slope ratio above the platform was 1:1.5, and the slope ratio below the platform was 1:1.75. The slope toe extended 0.675H to weaken the influence of the boundary range on the calculation results. Since the lower part of the solidified layer was rock, the compression deformation could be ignored. Therefore, the model boundary conditions were set as follows: XY - direction constraints were adopted on all four sides, the top was free, and 3 - direction constraints were adopted at the bottom. The constitutive model used was the Mohr - Coulomb model. A calculation model was established according to the above parameters as Figure 2 .

[0126] Input the calculation parameters in Table 6 above into FLAC3D, and sequentially establish finite - element calculation models based on three solidification construction methods (as Figures 3 to 6 ), and at the same time, for the shallow solidification double - layer foundation method and the shallow solidification + plain concrete pile composite foundation method, the pile spacing was determined by the user in combination with the actual situation on site and local codes. In this example, the pile spacing of the plain concrete pile composite foundation method was 2 m, and the pile spacing of the cement mixing pile composite foundation method was 1.5 m.

[0127] Thus, the obtained model structure was symmetric, so a half - structure was taken for analysis. Assume that the 10 - m embankment was filled in 4 times, that is, 2.5 m was filled each time, as Figure 3As shown in the figure, monitoring points are arranged at the center of the top of the subgrade, the upper interface of the solidified layer, the lower interface of the solidified layer, and at positions 16 m on both sides of the center line on the left and right. The settlement and earth pressure of the subgrade are monitored when the filling height reaches 2.5 m, 5 m, 7.5 m, and 10 m.

[0128] The maximum settlement values of the subgrade under different filling heights of the foundation under the conditions of three solidification methods are calculated and returned to the high-fill subgrade deep soft foundation improvement and solidification device for comparison and analysis. At the 10 m monitoring position, the maximum settlement values of the subgrade simulated by the unimproved and solidified foundation and the 7-day shallow solidification double-layer foundation method are 34.60 m and 30.05 m respectively. According to the "Highway Subgrade Design Code", the post-construction settlement value of the soft soil subgrade should meet the requirements of Table 8. However, it is obvious that the post-construction settlement values of the unimproved and solidified foundation and the 7-day shallow solidification double-layer foundation method are > 30 cm, which do not meet the code requirements. While the shallow solidification + plain concrete pile composite foundation method and the shallow solidification + cement mixing pile composite foundation method both meet the requirements. Considering the project cost, the shallow solidification + plain concrete pile composite foundation method is finally determined for this area.

[0129] In this step, the construction method of the second prediction model is similar to that of the first prediction model, including:

[0130] Obtain the solidification performance information, where the solidification performance information includes soil parameters, confining pressure, soil compressive strength and soil elastic modulus in the early and late stages of solidification;

[0131] Using the soil parameters and confining pressure as inputs, and the soil compressive strength and soil elastic modulus as outputs, construct a prediction model and evaluate the prediction accuracy of the prediction model;

[0132] If the prediction accuracy of the prediction model is lower than the preset expected accuracy, record the architecture of the prediction model, reconstruct the prediction model using the solidification performance information, and mask the recorded architecture until the prediction accuracy of the constructed prediction model reaches the expected accuracy to obtain the second prediction model.

[0133] Embodiment 2

[0134] The present application also provides a design system for the improvement and solidification plan of subgrade soil, including:

[0135] The first acquisition module is used to acquire soil solidification data. Each piece of the soil solidification data includes soil parameters, the curing agent formula used, and the soil compressive strength; through deep learning, the soil solidification data is expanded to obtain a curing agent information database;

[0136] A second acquisition module, configured to acquire soil parameters of a construction area, and screen in a curing agent information database according to the soil parameters to obtain a basic formula and corresponding strength data, where the basic formula includes raw materials in the formula and corresponding contents, and the strength data includes test strength data and predicted strength data;

[0137] A construction module, configured to construct an initial relationship equation based on the basic formula, with the soil curing strength as the response value and the contents of each raw material as influencing factors;

[0138] A design module, configured to perform coded level design of influencing factors through central composite design based on the basic formula and obtain corresponding true values, where the true values represent the contents of each raw material, and search for corresponding strength data according to the true values;

[0139] A fitting module, configured to fit the initial relationship equation according to the true values and strength data and perform variance analysis, and perform elimination processing on the terms in the initial relationship equation based on the variance analysis results to obtain an early strength equation and a late strength equation;

[0140] A first calculation module, configured to take the maximization of the soil curing strength as the goal, find the intersection of the solutions of the early strength equation and the late strength equation to obtain a target formula, and obtain an improved curing scheme according to the target formula.

[0141] As an optional implementation manner, the first acquisition module includes:

[0142] A classification unit, configured to traverse the soil curing data, classify the data with the same soil parameters and the same curing agent raw materials, and obtain a soil curing data packet for each type of soil;

[0143] A first construction unit, configured to perform deep learning based on each soil curing data packet, with the soil parameters and the curing agent formula as inputs and the soil compressive strength as the output, and construct a first prediction model;

[0144] An expansion unit, configured to multiply the soil parameters and the curing agent formula by different preset coefficients to obtain expansion parameters, and perform permutation and combination on the expansion parameters in sequence to obtain an expansion parameter group, and input the expansion parameter group into the first prediction model to obtain a predicted soil compressive strength;

[0145] A second construction unit, configured to store all expansion parameter groups and corresponding predicted soil compressive strengths, and construct a curing agent information database.

[0146] As an optional implementation manner, the second acquisition module includes:

[0147] The first acquisition unit is configured to acquire the material information within a preset range of the construction area and determine the priority raw materials;

[0148] The first screening unit is configured to screen in the curing agent information database based on the soil parameters to obtain the initially screened curing agent formulation data;

[0149] The second screening unit is configured to screen out the curing agent formulation data including the priority raw materials from the initially screened curing agent formulation data, and then screen the curing agent formulation data according to the corresponding soil compressive strength to obtain the basic formulation and the corresponding strength data.

[0150] As an optional implementation manner, it further includes:

[0151] The prediction module is configured to obtain the curing performance information based on the target formulation through a pre-established second prediction model, where the curing performance information includes the early curing strength, early elastic modulus, late curing strength, and late elastic modulus of the soil under different confining pressure conditions;

[0152] The second calculation module is configured to obtain the cohesion parameter and internal friction angle parameter of the soil after curing based on the curing performance information;

[0153] The simulation module is configured to construct a three-dimensional model of the foundation of the construction area based on the curing performance information, cohesion parameter, and internal friction angle parameter, and obtain the settlement displacement of the foundation under different curing construction methods through numerical simulation, and determine the target curing construction method based on the settlement displacement;

[0154] The generation module is configured to obtain the improved curing scheme according to the target formulation and the target curing construction method.

[0155] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all of them should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A method for designing an improved solidification scheme for roadbed soil, characterized in that: include: Acquire soil solidification data, each piece of soil solidification data includes soil parameters, a formula of a used solidifying agent, and soil compressive strength; Expanding the soil solidification data through deep learning to obtain a solidifying agent information database; Obtain soil parameters in the construction area, and screen in a curing agent information database according to the soil parameters to obtain a basic formula and corresponding strength data, wherein the basic formula includes raw materials in the formula and corresponding contents, and the strength data includes test strength data and predicted strength data; Based on the basic formula, the initial relationship equation is constructed with the soil solidification strength as the response value and the content of each raw material as the influencing factor; Based on the basic formula, a coding level design of the influencing factors is performed through a central composite design and a corresponding true value is obtained, wherein the true value represents the content of each raw material, and the corresponding intensity data is searched according to the true value; The initial relationship equation is fitted according to the real value and strength data and variance analysis is performed. Based on the variance analysis results, the items in the initial relationship equation are eliminated to obtain the early strength equation and the late strength equation. With the goal of maximizing the solidification strength of the soil, the intersection of the solutions of the early strength equation and the late strength equation is obtained to obtain a target formula, and an improved solidification scheme is obtained based on the target formula.

2. The improved solidification scheme design method for roadbed soil according to claim 1 is characterized in that: The soil solidification data is expanded by deep learning to obtain a solidifying agent information database, including: The soil solidification data are traversed, and data with the same soil parameters and the same curing agent raw materials are classified to obtain a soil solidification data packet for each soil; Based on each of the soil solidification data packets, deep learning is performed, with soil parameters and a curing agent formula as input and soil compressive strength as output, to construct a first prediction model; Multiplying soil parameters and curing agent formula by different preset coefficients to obtain extended parameters, and sequentially arranging and combining the extended parameters to obtain an extended parameter group, and inputting the extended parameter group into the first prediction model to obtain the predicted soil compressive strength; All extended parameter groups and the corresponding predicted soil compressive strength are stored to construct a curing agent information database.

3. The improved solidification scheme design method for roadbed soil according to claim 1, characterized in that: The step of obtaining soil parameters in the construction area and screening the curing agent information database according to the soil parameters to obtain a basic formula and corresponding strength data includes: Obtain material information within the preset range of the construction area and determine the priority raw materials; Based on the soil parameters, the curing agent information database is screened to obtain the initial screening curing agent formula data; The curing agent formula data including the priority raw materials are screened out from the initial screening curing agent formula data, and then the curing agent formula data are screened according to the corresponding soil compressive strength to obtain the basic formula and the corresponding strength data.

4. The improved solidification scheme design method for roadbed soil according to claim 1 is characterized in that: The method further comprises: Based on the target formula, solidification performance information is obtained through a pre-established second prediction model, wherein the solidification performance information includes early solidification strength, early elastic modulus, late solidification strength and late elastic modulus of the soil under different confining pressure conditions; Based on the solidification performance information, the cohesion parameters and internal friction angle parameters of the solidified soil are obtained; Based on the curing performance information, cohesion parameters and internal friction angle parameters, a three-dimensional model of the foundation in the construction area is constructed, and the settlement displacement of the foundation under different curing construction methods is obtained through numerical simulation, and a target curing construction method is determined based on the settlement displacement; An improved curing scheme is obtained according to the target formula and the target curing construction method.

5. The improved solidification scheme design method for roadbed soil according to claim 4, characterized in that: The method for constructing the second prediction model includes: Acquiring solidification performance information, wherein the solidification performance information includes soil parameters, confining pressure, soil compressive strength at early and late stages of solidification, and soil elastic modulus; With soil parameters and confining pressure as input and soil compressive strength and soil elastic modulus as output, a prediction model is constructed and the prediction accuracy of the prediction model is evaluated; If the prediction accuracy of the prediction model is lower than the preset expected accuracy, the architecture of the prediction model is recorded, the prediction model is rebuilt using the solidified performance information, and the recorded architecture is shielded until the prediction accuracy of the constructed prediction model reaches the expected accuracy, thereby obtaining a second prediction model.

6. The improved solidification scheme design method for roadbed soil according to claim 1, characterized in that: Based on the basic formula, the coding level design of the influencing factors is performed through a central composite design and the corresponding true values ​​are obtained, including: Determine the number of influencing factors of the central composite design based on the number of raw materials in the basic formula; According to the content range of each raw material in the basic formula, determine the central value, high value and low value of each influencing factor; Determine the level change step of each influencing factor according to the preset coding level and the high and low values ​​of the influencing factor; According to the coding level, level change step and central value of each influencing factor, the true value corresponding to the coding level is calculated.

7. A system for designing improved solidification schemes for roadbed soil, characterized in that: include: A first acquisition module is used to acquire soil solidification data, each piece of soil solidification data includes soil parameters, a formula of a used curing agent and soil compressive strength; Expanding the soil solidification data through deep learning to obtain a solidifying agent information database; The second acquisition module is used to obtain soil parameters in the construction area, and screen the curing agent information database according to the soil parameters to obtain a basic formula and corresponding strength data, wherein the basic formula includes raw materials in the formula and corresponding contents, and the strength data includes test strength data and predicted strength data; A construction module is used to construct an initial relationship equation based on the basic formula, with the soil solidification strength as the response value and the content of each raw material as the influencing factor; A design module, for performing a coding level design of the influencing factors through a central composite design based on the basic formula and obtaining corresponding true values, wherein the true values ​​represent the content of each raw material, and searching for corresponding intensity data according to the true values; A fitting module is used to fit the initial relationship equation according to the real value and the strength data and perform variance analysis, and eliminate the items in the initial relationship equation based on the variance analysis result to obtain the early strength equation and the late strength equation; The first calculation module is used to obtain the intersection of the solutions of the early strength equation and the late strength equation with the goal of maximizing the solidification strength of the soil body, obtain a target formula, and obtain an improved solidification plan based on the target formula.

8. The improved solidification scheme design system for roadbed soil according to claim 7, characterized in that: The first acquisition module includes: A classification unit, used for traversing the soil solidification data, classifying the data with the same soil parameters and the same curing agent raw materials, and obtaining a soil solidification data packet for each soil; A first construction unit is used to perform deep learning based on each of the soil solidification data packets, take soil parameters and a curing agent formula as input, and take soil compressive strength as output, to construct a first prediction model; An expansion unit is used to multiply soil parameters and curing agent formula by different preset coefficients to obtain expansion parameters, and sequentially arrange and combine the expansion parameters to obtain an expansion parameter group, and input the expansion parameter group into the first prediction model to obtain the predicted soil compressive strength; The second construction unit is used to store all the extended parameter groups and the corresponding predicted soil compressive strength to construct a curing agent information database.

9. The improved solidification scheme design system for roadbed soil according to claim 7, characterized in that: The second acquisition module includes: The first acquisition unit is used to acquire material information within a preset range of the construction area and determine the priority raw materials; The first screening unit is used to screen in the curing agent information database based on soil parameters to obtain the initial screening curing agent formula data; The second screening unit is used to screen out the curing agent formula data including the priority raw materials from the primary screening curing agent formula data, and then screen the curing agent formula data according to the corresponding soil compressive strength to obtain the basic formula and the corresponding strength data.

10. The improved solidification scheme design system for roadbed soil according to claim 7, characterized in that: Also includes: A prediction module, for obtaining curing performance information based on the target formula through a pre-established second prediction model, wherein the curing performance information includes early curing strength, early elastic modulus, late curing strength and late elastic modulus of the soil under different confining pressure conditions; The second calculation module is used to obtain the cohesion parameters and internal friction angle parameters of the soil after solidification based on the solidification performance information; A simulation module is used to construct a three-dimensional model of the foundation in the construction area based on the curing performance information, cohesion parameters and internal friction angle parameters, and obtain the settlement displacement of the foundation under different curing construction methods through numerical simulation, and determine the target curing construction method based on the settlement displacement; A generation module is used to obtain an improved curing scheme according to the target formula and the target curing construction method.

Images