Soil reinforcement effect evaluation method and system based on combination of machine learning and finite element

Abstract

The invention discloses a machine learning and finite element combined soil body reinforcement effect evaluation method and system. The method comprises the steps of obtaining data sets of reinforced soil bodies in different maintenance periods; constructing a soil mechanical parameter prediction model based on the wave velocity through a machine learning method, carrying out model training on the soil mechanical parameter prediction model according to the data set, and packaging the trained model into a prediction tool; after on-site grouting reinforcement construction, wave velocity data of a reinforced soil body are obtained through an on-site wave velocity test, the distribution range of the reinforced soil body is determined, and key mechanical parameters of the reinforced soil body are obtained based on a prediction tool; carrying out site construction process simulation through finite element modeling to obtain a deformation calculation result; and the deformation calculation result is compared with the deformation control standard of the adjacent underground structure, and the deformation control effect of grouting reinforcement is evaluated. The method solves the problem that the quantitative relation between the wave velocity and the mechanical parameters cannot be established in the traditional technology, and achieves the precise pre-judgment of construction on the deformation of the underground structure.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of soil reinforcement, and particularly relates to a soil reinforcement effect evaluation method and system based on machine learning combined with finite elements. BACKGROUND

[0002] With the continuous expansion of urban underground engineering construction scale, grouting reinforcement technology as the core technology of ensuring the safety of underground engineering near construction, its reinforcement effect is directly related to the stability and safety of underground structure. The traditional detection method mainly relies on the test after drilling core to evaluate the reinforcement effect. This evaluation method consumes more manpower and material resources, and also has a certain time lag. The randomness of sampling position also leads to insufficient spatial representativeness of the evaluation result. In addition, the traditional method is also difficult to evaluate whether the deformation of the structure near the reinforcement meets the control requirements. Therefore, it is of great significance to establish a simple and fast soil reinforcement effect evaluation method to ensure the safety of the adjacent underground structure, reduce the construction period and reduce the construction cost.

[0003] The patent document with the publication number CN119198901A discloses a method, system, medium and equipment for detecting the distribution of grout after grouting of rock-soil mass, which includes the following steps: before and after grouting, draw wave velocity contour lines in three-dimensional space for the grouting area; compare the wave velocity contour lines before and after grouting, obtain the wave velocity anomaly area, and then form a grout three-dimensional distribution model according to the wave velocity anomaly area inversion; wherein, the drawing step of wave velocity contour lines in three-dimensional space includes: calculating the wave velocity through the coordinate difference and time difference between the transmitter and receiver, establishing a wave velocity three-dimensional vector, discretizing the grouting area into equidistant wave velocity points, taking the distance from the wave velocity point to the straight line where the wave velocity three-dimensional vector is located as the weighting factor, weighting and distributing the wave velocities in different directions to obtain the corresponding wave velocity of each wave velocity point, and drawing the wave velocity contour lines in three-dimensional space. It can realize three-dimensional imaging detection of grouting effect, and has the advantages of wide coverage area, high detection precision, etc. However, this technology can only judge the grout diffusion range, and cannot establish a quantitative relationship between the wave velocity parameter and the mechanical parameter, and thus cannot evaluate the influence of construction after reinforcement on the deformation of the adjacent underground structure.

[0004] The patent document with the publication number CN117988321A discloses a loess bio-grouting reinforcement intelligent analysis system based on experiments and machine learning, which includes the following steps: performing soil property testing to obtain soil property parameters; preparing different proportions of bio-grouting solutions from different varieties of soybeans; bio-grouting reinforcement, indoor geotechnical experiment testing of loess engineering property parameters after processing; selecting loess soil property parameters and bio-grouting solution parameters as input features, and loess engineering property parameters after bio-grouting reinforcement as output labels to establish a loess bio-grouting reinforcement effect prediction model; and calling the trained machine learning model to predict the reinforcement effect of newly collected loess sample data. This method can predict the indoor test parameters of the reinforced soil based on machine learning, avoiding the frequent trial-and-error process in traditional methods. However, in actual engineering, after soil reinforcement, it is necessary to evaluate whether the deformation of construction near the building meets the deformation control requirements to ensure the safety of the adjacent building. The evaluation of the reinforcement effect is not limited to the indoor unit test of the reinforced soil, but also needs to be based on the site soil reinforcement, the specific characteristics of the adjacent structure, and the specific construction process to conduct a more comprehensive evaluation of the reinforcement effect. However, this technology lacks the above evaluation considerations.

[0005] Therefore, the present application is proposed. SUMMARY

[0006] The technical problem to be solved by the present application is that the prior art can only determine the diffusion range of the grout through wave velocity, cannot establish a quantitative relationship between wave velocity parameters and mechanical parameters, and cannot achieve accurate prediction of the deformation of underground structures during construction. The purpose is to provide a soil reinforcement effect evaluation method and system based on machine learning and finite elements, which determines the mechanical parameters of the soil through machine learning methods and wave velocity testing methods, and then establishes a finite element model to simulate the deformation of underground structures during adjacent construction and evaluate the reinforcement effect of the soil. The present application realizes direct prediction of mechanical parameters from wave velocity test data, solves the problem that traditional technologies cannot establish a quantitative relationship between wave velocity and mechanical parameters, and realizes accurate prediction of the deformation of underground structures during construction.

[0007] The present application is realized by the following technical solutions: In a first aspect, the present application provides a soil reinforcement effect evaluation method based on machine learning and finite elements, which includes the following steps: Obtaining a data set of reinforced soil at different curing periods; the data set is a data set of wave velocity and mechanical parameters changing with time; Constructing a soil mechanical parameter prediction model based on wave velocity through a machine learning method, training the soil mechanical parameter prediction model according to the data set, and packaging the trained model as a prediction tool; After the in-situ grouting reinforcement construction, the wave velocity data of the reinforced soil body are obtained through the in-situ wave velocity test, and the distribution range of the reinforced soil body is determined; within the distribution range, the key mechanical parameters of the reinforced soil body are obtained based on the prediction tool; According to the key mechanical parameters, the deformation calculation results are obtained through the in-situ construction process simulation by finite element modeling; the deformation calculation results are compared with the deformation control standard of the adjacent underground structure, and the deformation control effect of the grouting reinforcement is evaluated.

[0008] Further, the data set of the reinforced soil body in different maintenance periods is obtained, including: The basic physical and mechanical parameters of the in-situ soil body are obtained by testing the basic physical and mechanical parameters of the foundation through taking in-situ soil samples; According to the basic physical and mechanical parameters of the in-situ soil body and the design requirements, the slurry ratio of the soil body is determined; according to the slurry ratio of the soil body, the soil sample is stirred and mixed and loaded into a maintenance tank for different period maintenance, and the standard mechanical test sample of the reinforced soil body is prepared; The wave velocity test and mechanical test are carried out on the standard mechanical test sample in different maintenance periods, the wave velocity data, elastic modulus and shear strength data of the reinforced soil body under the corresponding maintenance period are obtained, and the data set of the reinforced soil body in different maintenance periods is established.

[0009] Further, the basic physical and mechanical parameters include water content, unit weight, specific gravity, void ratio, compression coefficient and shear strength.

[0010] Further, the soil mechanical parameter prediction model refers to a nonlinear mapping relationship between wave velocity and elastic modulus, cohesion and internal friction angle constructed by a machine learning method.

[0011] Further, the soil mechanical parameter prediction model based on wave velocity is constructed by a machine learning method, the soil mechanical parameter prediction model is trained according to the data set, and the trained model is packaged as a prediction tool, including: The wave velocity data are processed by feature engineering, a multi-dimensional feature matrix is constructed by introducing the quadratic term and logarithmic term of the wave velocity; The multi-dimensional feature matrix is normalized by a standardization method, and the normalized data set is divided into a training set and a test set in a ratio of 8:2; For the three mechanical parameters of elastic modulus, cohesion and internal friction angle, a neural network regression model with a single hidden layer is constructed respectively; and the neural network regression model weight is optimized by an adaptive gradient descent algorithm, and the maximum iteration number is set; According to the data set, the neural network regression model is trained, and the coefficient of determination and root mean square error are calculated to evaluate the model precision by using the test set; The trained and tested model is packaged as a prediction tool, and the input of the prediction tool is wave velocity, and the output is corresponding soil mechanical parameters.

[0012] Further, the field construction process simulation is performed by finite element modeling, including: According to the field stratum condition, construction design drawing and adjacent building structure drawing, a geometric model of the stratum, reinforced soil and adjacent structure is established in the finite element software; After the field soil reinforcement meets the time requirement, the field wave velocity test is used to obtain the shear wave velocity of the reinforced soil; the shear wave velocity of the reinforced soil is input into the prediction tool to obtain the key mechanical parameters of the reinforced soil; according to the field investigation report and the concrete grade of the adjacent structure, the stratum material parameters of the field stratum and the adjacent structure are obtained; the stratum material parameters, reinforced soil material parameters and structure material parameters are assigned to the corresponding geometric model; The established geometric model is meshed, and then the load is applied, the boundary conditions are set, and the finite element model of the project is established; and the field construction process is simulated by the finite element model.

[0013] Further, the method further includes: By arranging displacement monitoring points on the protected object, the deformation of the protected object during construction is monitored to obtain field monitoring data; the field monitoring data is used to verify the soil mechanical parameter prediction model and further optimize the model.

[0014] In a second aspect, the present application further provides a soil reinforcement effect evaluation system combining machine learning and finite element, which comprises: An acquisition unit is configured to acquire a data set of reinforced soil at different maintenance periods; the data set is a data set of wave velocity and mechanical parameters changing with time; A machine learning model and a training unit are configured to construct a soil mechanical parameter prediction model based on wave velocity by a machine learning method, to perform model training on the soil mechanical parameter prediction model according to the data set, and to package the trained model as a prediction tool; A model prediction unit is configured to, after field grouting reinforcement construction, acquire wave velocity data of the reinforced soil by field wave velocity test and determine the distribution range of the reinforced soil; within the distribution range, the key mechanical parameters of the reinforced soil are acquired based on the prediction tool; A finite element construction simulation unit is configured to simulate the field construction process by finite element modeling according to the key mechanical parameters to obtain deformation calculation results; A reinforcement effect evaluation unit is configured to compare the deformation calculation results with the deformation control standard of the adjacent underground structure to evaluate the deformation control effect of grouting reinforcement.

[0015] In a third aspect, the present application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the machine learning combined with finite element soil reinforcement effect evaluation method when executing the computer program.

[0016] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the machine learning combined with finite element soil reinforcement effect evaluation method.

[0017] Compared with the prior art, the present application has the following advantages and beneficial effects: 1. The machine learning combined with finite element soil reinforcement effect evaluation method and system, by introducing a machine learning method, a nonlinear quantitative relationship between wave velocity parameters and key mechanical parameters of soil is constructed. The conventional technology can only determine the range of slurry diffusion through wave velocity, while the present application realizes direct prediction from wave velocity test data to mechanical parameters through multi-dimensional feature engineering and neural network model, solving the problem that the conventional technology cannot establish a quantitative relationship between wave velocity and mechanical parameters.

[0018] 2. The machine learning combined with finite element soil reinforcement effect evaluation method and system, the machine learning prediction result is combined with finite element construction simulation to form a complete chain of "wave velocity test→ parameter prediction→ deformation simulation→ effect evaluation", and the model is continuously iterated and optimized using field data. This closed-loop system not only realizes accurate prediction of the deformation of underground structures during construction, but also dynamically improves the generalization ability and engineering applicability of the model by accumulating data from multiple engineering cases, significantly reducing the trial and error cost.

[0019] 3. The machine learning combined with finite element soil reinforcement effect evaluation method and system, the parameters of the reinforced soil can be conveniently and quickly obtained through field wave velocity testing, reducing the field drilling and sampling and testing links, solving the problem of insufficient spatial representativeness of the evaluation result caused by random sampling position, reducing the construction cost, shortening the construction period, and ensuring the construction and operation safety of underground structures. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which are included to provide a further understanding of the embodiments of the present application and constitute a part of the application, do not constitute a limitation to the embodiments of the present application. In the drawings: Figure 1 A flowchart of the machine learning combined with finite element soil reinforcement effect evaluation method; Figure 2 A specific evaluation flowchart of the embodiment 2 of the present application; Figure 3This is a flowchart illustrating the construction of the soil mechanical parameter prediction model based on wave velocity in Embodiment 2 of the present invention. Figure 4 This is a geometric model diagram of the foundation pit and subway in Embodiment 2 of the present invention; Figure 5 The above represents the calculation results of the unreinforced vertical deformation in Embodiment 2 of the present invention. Figure 6 The calculation results of vertical deformation for grouting reinforcement in Embodiment 2 of the present invention are shown. Figure 7 This is a schematic diagram of tunnel displacement monitoring according to Embodiment 2 of the present invention; Figure 8 This is a structural block diagram of a soil reinforcement effect evaluation system based on machine learning and finite element method according to the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0022] This invention proposes a method and system for evaluating the effectiveness of soil reinforcement using machine learning combined with finite element method. By taking soil samples from the field, and based on the soil's fundamental physical and mechanical parameters and design requirements, the grout mix ratio is determined. Reinforced soil samples are prepared and cured for different periods. Wave velocity and mechanical tests are conducted on the samples at different curing periods to obtain the wave velocity and mechanical parameters of the reinforced soil at each curing period. Furthermore, a wave velocity-based soil mechanical parameter prediction model is established using machine learning. After on-site grouting reinforcement, wave velocity data of the reinforced soil is obtained through on-site wave velocity tests to determine the distribution range of the reinforced soil. The machine learning model then provides the key mechanical parameters of the reinforced soil. Finally, finite element method modeling is used to simulate the construction process, analyze the impact of the construction process on adjacent underground structures, and evaluate the deformation control effect of the grouting reinforcement.

[0023] This invention introduces machine learning and finite element method to establish a convenient evaluation method for the effect of soil grouting reinforcement based on wave velocity testing. It can take into account different reinforcement cycles, the specific structural forms of adjacent buildings, and the construction process, reducing the on-site sampling process after grouting reinforcement. This provides effective technical support for shortening the construction cycle, reducing construction costs, and ensuring the safe construction and operation of underground projects.

[0024] Example 1 like Figure 1 As shown, this invention provides a method for evaluating the effectiveness of soil reinforcement using a combination of machine learning and finite element analysis. The method includes: Step 1: Obtain datasets of soils reinforced with different curing cycles; In this embodiment, step 1 specifically includes: Step 11, test the basic physical and mechanical parameters of the foundation by taking field soil samples, and obtain the basic physical and mechanical parameters of the field soil; Specifically, according to the design requirements of the project, during the geological exploration stage, field soil samples are taken by drilling, and the basic physical and mechanical parameters of the field soil samples, such as moisture content, unit weight, specific gravity, void ratio, compression coefficient, and shear strength, are tested to establish the basic physical and mechanical parameters of the field soil and form a basic data set.

[0025] Step 12, according to the basic physical and mechanical parameters of the field soil and the design requirements, determine the slurry ratio of the soil; according to the slurry ratio of the soil, mix and mix the retrieved soil samples and load them into the curing tank for different period of curing to prepare standard mechanical test samples of the reinforced soil; Specifically, according to the basic physical and mechanical parameters of the field soil and the design requirements (i.e. the designed construction process), determine the slurry ratio of the soil; according to the slurry ratio of the soil, mix and mix the retrieved soil samples, load the soil samples into the curing tank, cover the geotextile to prevent water evaporation; and cure the reinforced soil for different periods (including 7 days, 14 days, 21 days, 28 days, 42 days), drill the soil samples at the corresponding curing period, and prepare standard mechanical test samples of the reinforced soil.

[0026] Step 13, test the wave speed and mechanical properties of the standard mechanical test samples at different curing periods, obtain the wave speed data, elastic modulus and shear strength data of the reinforced soil at the corresponding curing period, and establish the data set of the reinforced soil at different curing periods; the data set is the data set of the wave speed and mechanical parameters changing with time; Step 2, construct a soil mechanical parameter prediction model based on wave speed by machine learning method, train the soil mechanical parameter prediction model according to the data set, and encapsulate the trained model as a prediction tool; In this embodiment, the soil mechanical parameter prediction model refers to the nonlinear mapping relationship between wave speed and elastic modulus, cohesion and internal friction angle constructed by machine learning method.

[0027] In this embodiment, step 2 specifically includes: Step 21, perform feature engineering processing on the wave speed data, and construct a multi-dimensional feature matrix by introducing the quadratic term ( ) and logarithmic term ( ) of wave speed to fully characterize the complex nonlinear relationship between wave speed and mechanical parameters.

[0028] Step 22, the multi-dimensional feature matrix is normalized by a standardized method to eliminate the influence of dimensional differences on model training, and the normalized data set is divided into a training set and a test set in a ratio of 8:2 to ensure the generalization ability of the model.

[0029] Step 23, for the three mechanical parameters of elastic modulus (E), cohesion (c) and internal friction angle (φ), a neural network regression model with a single hidden layer (15 neurons) is constructed respectively; and the neural network regression model weight is optimized by an adaptive gradient descent algorithm, and the maximum number of iterations is set to 2000 times to fully fit the training data.

[0030] Step 24, according to the data set, the neural network regression model is trained; after the model training is completed, the determination coefficient (R2) and the root mean square error (RMSE) are used to evaluate the model precision, and the reliability of the wave velocity prediction mechanical parameter is verified.

[0031] Step 25, the trained and tested model is packaged as a prediction tool, and the input of the prediction tool is the wave velocity and the output is the corresponding soil mechanical parameter. This provides parameter support for subsequent finite element modeling to evaluate whether the deformation of the construction process meets the deformation control requirements.

[0032] Step 3, after the on-site grouting reinforcement construction is completed, the wave velocity data of the reinforced soil is obtained through the on-site wave velocity test, and the distribution range of the reinforced soil is determined; within the distribution range, the key mechanical parameters of the reinforced soil are obtained based on the prediction tool; In this embodiment, after the on-site grouting reinforcement construction is completed, the wave velocity test can be performed on the on-site grouting reinforced soil at different ages (i.e. periods), the wave velocity data of the reinforced soil is obtained, the wave velocity data is processed, and the geometric distribution range of the reinforced soil is obtained; the cohesion and internal friction angle and other key mechanical parameters of the reinforced soil are obtained based on the trained soil mechanical parameter prediction model (i.e. prediction tool); Step 4, according to the key mechanical parameters, the on-site construction process is simulated by finite element modeling to obtain deformation calculation results; the deformation calculation results are compared with the deformation control standards of the adjacent underground structure to evaluate the deformation control effect of the grouting reinforcement.

[0033] In this embodiment, step 4 specifically includes: Step 41, according to the design scheme and key mechanical parameters of the project, a finite element model simulating the construction process is constructed, and displacement and other deformation calculation results of the adjacent underground structure are obtained by finite element calculation; ​Step 42, compare the deformation calculation result with the deformation control standard of the adjacent underground structure, evaluate the deformation control effect of grouting reinforcement: according to the comparison, judge whether it meets the requirements, and if the deformation meets the requirements, the next step of construction can be carried out. If the deformation control standard is not met, multiple grouting, increasing support structure and other engineering measures can be taken.

[0034] Specifically, the field construction process simulation is carried out by finite element modeling, including: (1) According to the field stratum condition, construction design drawing and adjacent building structure drawing, the geometric model of stratum, reinforced soil and adjacent structure is established in the finite element software; (2) After the field soil reinforcement meets the time requirement, the shear wave velocity of the reinforced soil is obtained by field wave velocity test; the shear wave velocity of the reinforced soil is input into the prediction tool to obtain the key mechanical parameters of the reinforced soil; according to the field investigation report and the concrete grade of the adjacent structure, the stratum material parameters of the field stratum and the adjacent structure are obtained; the stratum material parameters, reinforced soil material parameters and structure material parameters (such as tunnel lining structure parameters) are assigned to the corresponding geometric model; (3) The established geometric model is meshed, then the load is applied, the boundary condition is set, and the finite element model of the project is established; and the field construction process is simulated by finite element model.

[0035] As a further implementation, the method further comprises: By laying displacement monitoring points on the protected object, the deformation of the protected object during construction is monitored to obtain field monitoring data; the field monitoring data is used to verify the soil mechanical parameter prediction model, and different engineering cases are used to accumulate data set to further optimize and improve the accuracy of the model.

[0036] Embodiment 2 As shown in Figures 2 to 7 , this embodiment is implemented according to the method steps of embodiment 1, as follows: (1) Project background A city subway tunnel construction has been put into operation, and now a deep foundation pit project is planned above the tunnel for the construction of an underground commercial complex due to the need of city development. The minimum vertical distance between the tunnel crown and the bottom of the foundation pit is 6.7 meters, and the deformation control requirement of the tunnel is 10 mm. Due to the large disturbance of the soil during the excavation of the foundation pit, it poses a serious threat to the structural stability and operation safety of the subway tunnel. Therefore, effective soil reinforcement measures must be taken to ensure the operation safety of the tunnel during the excavation of the foundation pit. Based on the geological conditions and engineering requirements, the soil grouting reinforcement method is used to systematically reinforce the soil around the tunnel. The main purpose of reinforcement is to improve the shear strength and bearing capacity of the soil around the tunnel, reduce the deformation of the subway tunnel caused by the excavation of the foundation pit, and prevent the deformation or damage of the tunnel structure. After the soil reinforcement on site, the reinforcement effect of the soil needs to be evaluated, and the influence of the foundation pit excavation on the deformation of the subway tunnel after the soil reinforcement is analyzed. Therefore, the following evaluation method is designed, and the specific evaluation process is as shown in Figure 2 .

[0037] (2) Field sampling and basic physical and mechanical parameter testing According to the results of the field geological investigation, the soil in the engineering influence range is mainly divided into 4 layers, the first layer is clay, the second layer is silt, the third layer is silty clay, and the fourth layer is clay. According to the design requirements, the reinforced soil is mainly the third layer of silty clay. The basic physical and mechanical properties of the soil sample are tested, and the test results are shown in Table 1.

[0038] Table 1 Test results of physical and mechanical parameters of soil

[0039] (3) Wave velocity and mechanical parameter testing of sample Through field sampling of the third layer of silty clay in the reinforced soil layer, according to the basic physical and mechanical properties of the soil sample and the design requirements, the cement parameter is 15%, the soil sample is fully stirred with cement slurry, and the stirred cement soil is filled into a curing tank. The curing tank is wrapped with geomembrane, and the corresponding soil sample is drilled to prepare standard triaxial samples at the corresponding curing age (i.e. period) of 7 days, 14 days, 21 days, 28 days and 42 days. Through the wave velocity testing instrument, the shear wave velocity value of the sample is tested, and after the wave velocity test is completed, the mechanical test of the sample is carried out to obtain the mechanical parameters of the sample. The data of each curing age includes 30 sample data, and the curing time, shear wave velocity, elastic modulus, cohesion and internal friction angle data set of the reinforced soil is established. The statistical results of the data set are shown in Table 2.

[0040] Table 2 Statistical results of data set of reinforced soil

[0041] (4) Wave velocity-based soil mechanical parameter prediction model Based on the correlation data of the shear wave velocity (Vs) and mechanical parameters (elastic modulus E, cohesion c, and internal friction angle φ) of the reinforced soil in Table 2, a wave velocity-driven mechanical parameter prediction model (i.e., a wave velocity-based soil mechanical parameter prediction model) was constructed using a machine learning method. First, a multi-dimensional feature matrix was constructed for the wave velocity data: taking the shear wave velocity Vs as the core feature, introducing its quadratic term ( ) and logarithmic term ( ), forming a feature combination containing nonlinear relationships to represent the complex mapping rule between wave velocity and mechanical parameters. Subsequently, the feature data was standardized to eliminate dimensional differences, and the data set was divided into a training set and a test set according to an 8:2 ratio. For the three target parameters of elastic modulus, cohesion, and internal friction angle, single-hidden-layer (with 15 neurons) fully connected neural network regression models were constructed, the adaptive gradient descent algorithm was used to optimize the network weights, and the maximum number of iterations was set to 2000 to ensure that the model was fully converged. The flowchart of model construction is shown in Figure 3 After the model training was completed, the coefficient of determination ( ) and the root mean square error (RMSE) were calculated to evaluate the performance based on the test set data. The results showed that the for elastic modulus prediction was 0.99, the cohesion was 0.96, and the internal friction angle was 0.85; the RMSE for elastic modulus prediction was 49.02, the cohesion was 2.34, and the internal friction angle was 1.13, verifying the model's prediction ability for mechanical parameters. The trained model was encapsulated as a prediction tool that could output mechanical parameters based on input wave velocity.

[0042] (5) In-situ reinforced soil wave velocity testing and mechanical parameter prediction After the completion of in-situ grouting reinforcement, according to the results of in-situ wave velocity testing after 28 days of reinforcement, the shear wave velocity of the reinforced soil was 550 m / s. The shear wave velocity of the reinforced soil was used as input to predict the elastic modulus, cohesion, and internal friction angle of the reinforced soil using the mechanical parameter prediction tool. The predicted results were an elastic modulus of 1849.4 MPa, a cohesion of 60.0 kPa, and an internal friction angle of .

[0043] (6) Finite element model construction and construction process simulation According to the design of the foundation pit and the positional relationship between the foundation pit and the subway tunnel, the geometric model diagram of the foundation pit and the subway was drawn, as shown in Figure 4As shown, the geometric model is imported into finite element software, and geometric models of the foundation pit and subway are established through stretching, sweeping, cutting, and moving. The constitutive model of the soil adopts the hardened soil constitutive model. The hardened soil model can reflect the difference in mechanical properties of soil under loading and unloading and has been widely used in numerical simulation of foundation pit excavation. In the hardened soil model, the stiffness parameter of the soil is the triaxial experimental secant stiffness. Tangential stiffness of the main compaction loading test Unloading elastic modulus Three parameters, assuming the unloading of the soil satisfies the linear elastic relationship, then the unloading elastic modulus... The elastic modulus can be determined by on-site wave velocity testing. The values ​​of the other two stiffness parameters are proportional to each other. The strength parameters are determined to be the results predicted by the mechanical parameter prediction tool. The model parameter values ​​for the soil are shown in Table 3.

[0044] Table 3 Soil Model Parameter Table

[0045] The geometric model was meshed, and corresponding material parameters were assigned to the mesh. Loads and boundary conditions were applied to the model to establish a finite element model of the project. The displacement of the subway tunnel caused by the excavation of the foundation pit before and after reinforcement was simulated using the finite element model. The calculation results are as follows: Figure 5 and Figure 6 As shown in the figure, according to the calculation results, without soil reinforcement during foundation pit excavation, the vertical displacement of the tunnel was 10.03 mm, which did not meet the deformation control requirements. After soil reinforcement, the maximum vertical displacement of the tunnel was 8 mm, which met the deformation control requirements. This indicates that soil reinforcement can effectively control the deformation of subway tunnels and ensure the safe operation of the subway.

[0046] (7) Verification of the effectiveness of information-based foundation pit excavation and evaluation Monitoring points were set up along the tunnel lining. During the excavation of the foundation pit, measurements were taken three times a day, morning, noon, and evening, to record the displacement values ​​of the tunnel. The layout of the monitoring points is shown in the diagram below. Figure 7 As shown in the diagram. If the displacement exceeds the warning value during excavation, measures are taken to increase the support structure, reduce the size of the excavation area, and, if necessary, increase the ballast load on the excavated area to prevent significant deformation of the subway. For the excavated area, the construction of the foundation pit slab and underground structure should be carried out as soon as possible, and the soil should be backfilled after the underground structure construction is completed. According to the results of on-site displacement monitoring, the maximum displacement of the subway tunnel during construction was 6 mm, which is close to the result calculated by the model, indicating that this method can effectively evaluate the soil reinforcement effect.

[0047] (8) Model optimization and improvement Through the collection of the data of the engineering case, the data set of similar engineering cases is established, the measured data of the engineering case is used for the training of the model, the prediction model is iteratively predicted, and the prediction accuracy of the model is improved and the application range of the model is expanded.

[0048] Embodiment 3 As Figure 8 shown, the difference between the present embodiment and embodiment 1 is that the present embodiment further provides a soil reinforcement effect evaluation system combining machine learning and finite element, which corresponds to the soil reinforcement effect evaluation method combining machine learning and finite element of embodiment 1; the system comprises: An acquisition unit is configured to acquire a data set of the reinforced soil in different maintenance periods; the data set is a data set of the wave velocity and the mechanical parameters changing with time; A machine learning model and a training unit are configured to construct a soil mechanical parameter prediction model based on the wave velocity by using a machine learning method, to perform model training on the soil mechanical parameter prediction model according to the data set, and to encapsulate the trained model as a prediction tool; A model prediction unit is configured to acquire the wave velocity data of the reinforced soil by performing a field wave velocity test after a field grouting reinforcement construction, and to determine the distribution range of the reinforced soil; within the distribution range, the key mechanical parameters of the reinforced soil are acquired based on the prediction tool; A finite element construction simulation unit is configured to simulate the field construction process by using finite element modeling according to the key mechanical parameters, and to acquire a deformation calculation result; A reinforcement effect evaluation unit is configured to compare the deformation calculation result with a deformation control standard of the adjacent underground structure, and to evaluate the deformation control effect of the grouting reinforcement.

[0049] The execution process of each unit can be performed according to the method flow steps of the soil reinforcement effect evaluation method combining machine learning and finite element of embodiment 1, and the details are not described herein.

[0050] Meanwhile, the present application further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned soil reinforcement effect evaluation method combining machine learning and finite element when executing the computer program.

[0051] Meanwhile, the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the above-mentioned soil reinforcement effect evaluation method combining machine learning and finite element.

[0052] Those skilled in the art will appreciate that embodiments of the application can be readily used as software, hardware, or a combination of software and hardware. In a software embodiment, the methods can be tangibly embodied in a machine-readable storage medium having stored thereon instructions that can be used to program a computer to perform any of the operations described herein. The software implementation can be initialized to execute via a processor to obtain features disclosed herein.

[0053] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing each of the flowchart blocks or steps

[0054] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 means for functionally implementing each of the flowchart blocks or steps

[0055] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flowcharts and / or blocks Figure 1 Figure 1 means for functionally implementing each of the flowchart blocks or steps

[0056] The above description is only specific embodiments of the present application, and is not intended to limit the protection scope of the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for evaluating soil reinforcement effect by machine learning combined with finite element, characterized in that, The method comprises: obtaining a data set of reinforced soil bodies in different maintenance periods; the data set is a data set of wave velocity and mechanical parameters changing with time; constructing a soil body mechanical parameter prediction model based on wave velocity by a machine learning method, training the soil body mechanical parameter prediction model according to the data set, and packaging the trained model as a prediction tool; after the in-situ grouting reinforcement construction, obtaining the wave velocity data of the reinforced soil body through the in-situ wave velocity test, and determining the distribution range of the reinforced soil body; within the distribution range, the key mechanical parameters of the reinforced soil body are obtained based on the prediction tool; according to the key mechanical parameters, simulating the in-situ construction process by finite element modeling to obtain deformation calculation results; comparing the deformation calculation results with the deformation control standard of the adjacent underground structure to evaluate the deformation control effect of grouting reinforcement.

2. The method for evaluating soil reinforcement effect of machine learning combined with finite element according to claim 1, characterized in that, The method comprises: obtaining a data set of reinforced soil bodies in different maintenance periods; the data set is a data set of wave velocity and mechanical parameters changing with time; obtaining the basic physical and mechanical parameters of the in-situ soil by testing the basic physical and mechanical parameters of the in-situ soil sample; determining the slurry ratio of the soil according to the basic physical and mechanical parameters of the in-situ soil and the design requirements; mixing and stirring the retrieved soil sample according to the slurry ratio of the soil, and loading it into a curing tank for different period maintenance to prepare standard mechanical test samples of the reinforced soil body; 3. The method of claim 2, wherein, performing wave velocity test and mechanical test on the standard mechanical test samples in different maintenance periods to obtain the wave velocity data, elastic modulus and shear strength data of the reinforced soil body in the corresponding maintenance period, and establish the data set of the reinforced soil body in different maintenance periods.

4. The method of claim 1, wherein, The basic physical and mechanical parameters include water content, specific gravity, bulk density, void ratio, compression coefficient and shear strength.

5. The method of claim 4, wherein the method further comprises: The soil body mechanical parameter prediction model refers to a nonlinear mapping relationship between wave velocity and elastic modulus, cohesion and internal friction angle constructed by a machine learning method. The method comprises: performing feature engineering processing on the wave velocity data to construct a multi-dimensional feature matrix by introducing the quadratic term and logarithmic term of the wave velocity; normalizing the multi-dimensional feature matrix by a standardization method, and dividing the normalized data set into a training set and a test set in a ratio of 8:2; constructing a neural network regression model with a single hidden layer for each of the elastic modulus, cohesion and internal friction angle; and optimizing the neural network regression model weight by an adaptive gradient descent algorithm, and setting the maximum number of iterations; training the neural network regression model according to the data set, and calculating the coefficient of determination and root mean square error to evaluate the model accuracy using the test set; 6. The method of claim 1, wherein, packaging the trained and tested model as a prediction tool, and the input of the prediction tool is wave velocity and the output is the corresponding soil body mechanical parameter. The method comprises: establishing a geometric model of the stratum, reinforced soil body and adjacent structure in the finite element software according to the in-situ stratum condition, construction design drawing and adjacent building structure drawing; After the on-site soil body reinforcement meets the time requirement, the shear wave velocity of the reinforced soil body is obtained by on-site wave velocity testing; the shear wave velocity of the reinforced soil body is input into the prediction tool to obtain the key mechanical parameters of the reinforced soil body; the stratum material parameters of the on-site stratum and the adjacent structure are obtained according to the on-site survey report and the concrete grade of the adjacent structure; the stratum material parameters, the reinforced soil body material parameters and the structure material parameters are assigned to the corresponding geometric model; The established geometric model is meshed, then the load is applied, the boundary conditions are set, and the finite element model of the project is established; and the on-site construction process is simulated by the finite element model.

7. The method of claim 1, wherein the method is characterized by: The method further comprises: The deformation of the protected object in the construction process is monitored by arranging displacement monitoring points on the protected object, and the on-site monitoring data is obtained; the on-site monitoring data is used to verify the soil body mechanical parameter prediction model and further optimize the model. 8.A soil reinforcement effect evaluation system using machine learning and finite elements, characterized by The system comprises: An acquisition unit is configured to acquire a data set of the reinforced soil body in different maintenance periods; the data set is a data set of the wave velocity and the mechanical parameters changing with time; A machine learning model and a training unit are configured to construct a soil body mechanical parameter prediction model based on wave velocity by a machine learning method, to perform model training on the soil body mechanical parameter prediction model according to the data set, and to encapsulate the trained model as a prediction tool; A model prediction unit is configured to acquire the wave velocity data of the reinforced soil body by on-site wave velocity testing after the on-site grouting reinforcement construction, and to determine the distribution range of the reinforced soil body; in the distribution range, the key mechanical parameters of the reinforced soil body are obtained based on the prediction tool; A finite element construction simulation unit is configured to simulate the on-site construction process by finite element modeling according to the key mechanical parameters, and to obtain a deformation calculation result; A reinforcement effect evaluation unit is configured to compare the deformation calculation result with the deformation control standard of the adjacent underground structure, and to evaluate the deformation control effect of the grouting reinforcement.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to realize the machine learning combined finite element soil body reinforcement effect evaluation method of any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1-9. The computer program is executed by the processor to realize the machine learning combined finite element soil body reinforcement effect evaluation method of any one of claims 1 to 7.

Images