Method and system for determining soil shear strength index based on random forest-based load test

Abstract

The application discloses a method and system for determining soil shear strength index based on a random forest, and the method comprises the following steps: acquiring actual engineering data and load test p-s curves and corresponding soil shear strength indexes simulated by software; extracting characteristic data p cr 、s cr 、a and b of the load test p-s curves, and establishing a training sample set; training a random forest regression model by taking the characteristic data of the load test in the training sample set as model input, and taking corresponding soil shear strength indexes c or phi as output; acquiring the load test p-s curves of actual engineering, extracting characteristic data, and inputting the characteristic data into the trained random forest regression model, and finally outputting the numerical values of the shear strength indexes c and phi of the soil measured by the load test of actual engineering. The method not only expands the engineering application range of the load test, but also provides a new method for obtaining the in-situ shear strength index of soil.

Description

TECHNICAL FIELD

[0001] The present application belongs to the cross technical field of computer application technology and in-situ testing technology of engineering investigation, and particularly relates to a method and system for determining soil shear strength indexes based on a random forest. BACKGROUND

[0002] Load test is a common in-situ testing technology of soil in engineering investigation, and appears in many engineering investigation specifications, foundation design specifications and engineering geology handbooks and other professional books, has a mature standard test procedure, and is often used to determine soil mechanical indexes such as foundation bearing capacity, soil deformation modulus, and foundation soil base reaction coefficient. At present, load test is only used to determine soil shear strength parameters in very limited cases, and the non-stable method (non-draining condition) load test is used to estimate the undrained shear strength (only index c) of saturated clay, and there is no report on the determination of soil shear strength (index c, ) by a conventional test method.

[0003] The shear strength index of soil is an important index for analyzing engineering geological stability problems such as foundation stability and slope stability. For soil that is difficult to sample or easy to be disturbed, the in-situ shear strength index obtained by in-situ testing method is the only reliable method to ensure the shear strength index c, . Therefore, it is of great practical significance to determine the soil shear strength by load test.

[0004] The load test data are mainly the load p borne by the pressure plate and the corresponding settlement s of the pressure plate when reaching stability, and the test results are often displayed by a p-s curve. Engineering practice and related research show that the typical p-s curve has significant morphological characteristics, which can be divided into three stages of linear deformation stage, local shear stage and failure stage, and the end point of the linear segment is the plastic load point p cr and the corresponding settlement s cr , and the curve after the linear segment on the p-s curve conforms to the hyperbolic law.

[0005] The plastic load point on the p-s curve of load test and the curve shape after the plastic load point are controlled by the shear strength of soil, so the shear strength parameters of soil except saturated clay can be determined from the load curve. However, due to the complex relationship between the p-s curve of load test and the shear strength of soil, it is impossible to analyze the relationship between them by using conventional multivariate statistical method. SUMMARY

[0006] The present application aims at the above-mentioned deficiencies of the prior art, and provides a method and system for determining soil shear strength indexes based on a random forest.

[0007] To achieve the above-mentioned purpose, the present application adopts the following technical solution:

[0008] The first objective of this invention is to provide a method for determining soil shear strength parameters based on random forest load tests, comprising the following specific steps:

[0009] S1. Obtain actual engineering data and load test curves and corresponding shear strength indicators from software simulation;

[0010] S2. Extract the feature data p of all load test ps curves from step S1. cr s cr a and b, and establish a training sample set, where p cr and s cr denoted as the bearing plate load and bearing plate settlement value corresponding to the end point of the straight line segment of the ps curve, and a and b are the fitting coefficients of the curve segment after the straight line segment of the ps curve.

[0011] S3. Train a random forest regression model using the training sample set, and use the load test feature data p from the training sample set. cr s cr a and b are used as model inputs, and the corresponding soil shear strength index c or For the output, train a random forest regression model, where c is the cohesion. It is the internal friction angle;

[0012] S4. Obtain the load test ps curve data of the soil being tested in the actual project, and extract the characteristic data p representing the load test ps curve of the soil being tested in the actual project. cr s cr The feature data, a and b, are input into a pre-trained random forest regression model. The random forest regression model ultimately outputs the shear strength index c of the soil tested in the actual engineering load test. The value.

[0013] Furthermore, in step S1, the corresponding shear strength index in the actual engineering data refers to the shear strength test index data of the soil measured by the actual engineering load test.

[0014] Furthermore, in step S1, the load test curve and corresponding shear strength index simulated by the software refer to the load test curve simulated using FLAC numerical simulation software and the shear strength index used in the simulation.

[0015] Furthermore, in step S2, the endpoint of the straight line segment of the ps curve is determined by the change in the slope of the ps curve being less than 0.5.

[0016] Furthermore, the change in the slope of the ps curve is determined by linear interpolation of the slope changes between two adjacent test points on the ps curve. The formula for calculating the change in the slope of a certain test point i on the ps curve is:

[0017]

[0018] in: s i p i These are the values ​​of s and p at measurement point i; s max p max These are the s and p values ​​corresponding to the last load level in the ps curve, where p is the load applied to the bearing plate and s is the settlement under the corresponding load.

[0019] Furthermore, in step S2, the feature data a and b are fitted using hyperbolic curves from the data points of the three load levels preceding the last load level in the load test. The fitting formula is as follows:

[0020] s / p = a·s + b,

[0021] Where p is the load applied to the bearing plate, and s is the settlement under the corresponding load.

[0022] Furthermore, in step S3, the random forest regression model is trained using the training sample set as follows: 80% of the training samples are randomly selected to train the random forest regression model, and the remaining 20% ​​of the samples are used to test the model.

[0023] Furthermore, in step S3, the parameters for setting the random forest regression model include: the maximum depth of each decision tree is set to 5, the number of random forest regression trees is set to 100, the minimum sample size for node splitting is 2, and the minimum sample size for leaf nodes is 1.

[0024] A second objective of this invention is to provide a system for determining soil shear strength parameters based on random forest load tests, the system comprising:

[0025] The data acquisition module is used to acquire load test PS curves and corresponding soil shear strength index test data from actual engineering data, as well as load test PS curves and corresponding soil shear strength indexes simulated by numerical simulation software.

[0026] The dataset construction module is used to extract the feature data p of all load test ps curves from the data acquisition module. cr s cr a and b, and establish a training sample set, where p cr and s crdenoted as the bearing plate load and bearing plate settlement value corresponding to the end point of the straight line segment of the ps curve, and a and b are the fitting coefficients of the curve segment after the straight line segment of the ps curve.

[0027] The data training module trains a random forest regression model using the sample set in the dataset construction module, and uses the load test feature data p from the training sample set. cr s cr a and b are used as model inputs, and soil shear strength index c or For the output, train a random forest regression model, where c is the cohesion. It is the internal friction angle;

[0028] The practical engineering application module acquires load test ps curve data of the tested soil in actual engineering projects and extracts characteristic data p representing the load test ps curve of the tested soil in actual engineering projects. cr s cr a and b, and input the feature data into the random forest regression model trained by the data training module, outputting the shear strength index c of the soil tested in the actual engineering load test. The value.

[0029] A third objective of this invention is to provide a computer-readable storage medium storing a computer program configured to, when invoked by a processor, implement the steps of the above-described method for determining soil shear strength indices using a random forest-based load test.

[0030] A fourth object of the present invention is to provide an apparatus for determining soil shear strength indices, the apparatus comprising at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the above-described method for determining soil shear strength indices using random forest-based load tests.

[0031] Compared with the prior art, the beneficial effects of the technical solution provided by the present invention are as follows:

[0032] (1) This invention provides a method and system for determining soil shear strength indices based on random forest load tests. The method first determines the characteristic data and extraction methods used to characterize the entire process of the load test ps curve; then, using data obtained from actual engineering surveys and numerical simulations of load tests, a random forest regression analysis sample set for determining soil shear strength based on load tests is established; finally, a system is constructed using the load test characteristic data as input and the soil shear strength index c or... The output is a random forest regression analysis model, which is then learned and validated using sample set data to form a random forest regression model that determines the soil shear strength index from load test data. Finally, in engineering practice, feature data can be extracted from the load test data and input into the random forest regression model to obtain the soil shear strength index c or...

[0033] (2) Using load tests to determine soil shear strength index expands the engineering application of load tests and provides a new method for obtaining soil in-situ shear strength index. Load tests are an essential test item for large-scale engineering construction projects. This invention effectively expands the application field of load tests, greatly improves the automation and accuracy of load test data processing, and also enriches the methods for obtaining soil in-situ shear strength index. It can save engineering survey costs and improve engineering survey accuracy, which is of great significance to engineering construction.

[0034] (3) This invention applies random forest to load test data processing, which is an important breakthrough in determining soil shear strength index. It provides an important technical means and proposes a standardized processing method for load test data. The soil shear strength index obtained by this method is true and reliable. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method for determining soil shear strength index based on random forest load tests according to the present invention;

[0036] Figure 2 This is a graph showing the fit between the load test ps curve of this invention and the ps curve obtained by backtracking using feature data;

[0037] Figure 3 This is a flowchart of the random forest regression model formation process of the present invention;

[0038] Figure 4 It is a scatter plot comparing the shear strength index c determined by the random forest regression model with the actual c, using the test sample set.

[0039] Figure 5 The test sample set is used to test the random forest regression model to determine the shear strength index. With reality A scatter plot for comparison. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the specific embodiments and accompanying drawings are described in further detail below. Where specific techniques or conditions are not specified in the embodiments, they are performed in accordance with the techniques or conditions described in the literature in this field or according to the product manual.

[0041] The technical terms used in this invention are explained as follows:

[0042] Load testing is an in-situ testing method for foundation soil, which can be used to determine the bearing capacity and deformation characteristics of soil and rock within the main influence range of bearing plate stress;

[0043] ps curve: p is the load applied to the bearing plate, and s is the settlement under the corresponding load;

[0044] p cr : The load on the bearing plate corresponding to the endpoint of the straight line segment of the ps curve;

[0045] s cr : The settlement value of the bearing plate corresponding to the end of the straight line segment of the ps curve;

[0046] a and b: Fitting coefficients for the straight line segment and the curve segment after the straight line segment of the ps curve;

[0047] The shear strength index c of soil refers to cohesion;

[0048] soil shear strength index The internal friction angle.

[0049] The range of physical and mechanical parameters of the soil in the load test numerical simulation provided by this invention was determined with reference to the "Engineering Geology Handbook". The bearing plate used has an area of ​​0.5m². 2 The circular rigid bearing plate, with the magnitude of each load level and the termination conditions of the simulation test all conforming to the provisions of Appendix C.0.5 of the "Code for Design of Building Foundations" (GB50007-2011). Using FLAC numerical simulation software, finite difference numerical models with different soil physical parameters were established. Simulation calculations were performed to obtain the deformation data of each load level and the stable condition under each load in the load test for each soil parameter model.

[0050] An embodiment of the first aspect of the present invention proposes a method for determining soil shear strength parameters based on random forest load tests, the basic process of which is as follows: Figure 1 As shown. The method includes:

[0051] Step S1: Obtain actual engineering data and the load test ps curve and corresponding soil shear strength index from software simulation;

[0052] Step S2: Extract the feature data p of all ps curves from step S1. cr s cr a and b, and establish a training sample set;

[0053] Step S3: Train a random forest regression model using the training sample set, and use the load test feature data p from the training sample set. cr s cra and b are used as model inputs, and the corresponding shear strength index c or For output, train a random forest regression model;

[0054] Step S4: Obtain the load test ps curve test data of the soil under test in the actual project, and extract the characteristic data p that characterizes the load test ps curve of the tested project. cr s cr The feature data, a and b, are input into a pre-trained random forest regression model. The random forest regression model ultimately outputs the shear strength index c of the soil tested in the actual engineering load test. The value.

[0055] First, determine the characteristic data and extraction methods used to characterize the entire process of the ps curve in the load test; then, using data obtained from actual engineering surveys and numerical simulations of load tests, establish a random forest regression analysis sample set for determining soil shear strength through load tests; finally, construct a system using the characteristic data from the load tests as input and the soil shear strength index c or The output is a random forest regression analysis model, which is then learned and validated using sample set data to form a random forest regression model that determines the soil shear strength index from load test data. Finally, in engineering practice, feature data can be extracted from the load test data and input into the random forest regression model to obtain the soil shear strength index c or... Applying random forests to load tests to determine soil shear strength parameters expands the engineering application scope of load tests and provides a new method for obtaining in-situ soil shear strength parameters.

[0056] It should be noted that in step S1, the load test PS curve in the actual engineering survey data refers to the load test PS curve obtained using conventional methods in this field for the actual engineering project. The corresponding soil shear strength index refers to the shear strength test index data of the soil measured by the load test in the actual engineering project, obtained using conventional methods in this field. The load test PS curve simulated by software is not specifically limited here. In this implementation, FLAC numerical simulation software is used to establish finite difference numerical models with different soil physical property parameters. The simulation calculations are used to obtain the load test PS curve and the shear strength index used in the simulation for each load level under each soil parameter model and the deformation data when the load is stable.

[0057] Specifically, the feature data extraction method in this embodiment can determine the endpoint of the straight line segment of the ps curve by the change value of the slope of the ps curve being less than 0.5. No specific limitation is made here. The change value of the slope of the ps curve is determined by linear interpolation of the slope changes of two adjacent test points on the ps curve. The formula for calculating the change value of the slope of a certain test point i on the ps curve is:

[0058]

[0059] in: s i p i These are the values ​​of s and p at measurement point i; s max p max These are the s and p values ​​corresponding to the last load level in this set of ps data, where p is the load applied to the bearing plate and s is the settlement under the corresponding load.

[0060] Feature data a and b are the fitting coefficients for the curve segment after the straight line segment of the ps curve. Data points from the three load levels preceding the last load level in the load test are used to fit a hyperbola. The fitting formula is:

[0061] s / p = a·s + b,

[0062] Where: p is the load applied to the bearing plate, and s is the settlement under the corresponding load.

[0063] It should be noted that the formulas and data point requirements for extracting feature data disclosed in this invention can be replaced with other formulas and data point requirements. Due to different experimental conditions or environments, the obtained ps curves may have certain deviations. Therefore, any formulas and corresponding feature data extraction methods obtained based on the logic of this invention should fall within the protection scope of this invention.

[0064] Using the feature data obtained above, a dataset is constructed to train a random forest regression model. No specific limitations are specified here. The load test feature data p from the training sample set is used. cr s cr a and b are used as model inputs, and soil shear strength index c or For the output, train a random forest regression model.

[0065] The specific steps for training a random forest regression model using the training sample set are as follows: 80% of the training samples are randomly selected to train the random forest regression model, and the remaining 20% ​​of the samples are used to test the model. The settings for the random forest regression model include: the maximum depth of each decision tree is set to 5, the number of random forest regression trees is set to 100, the minimum sample size for node splits is 2, and the minimum sample size for leaf nodes is 1.

[0066] It should be noted that the construction method of the random forest regression model is not limited here. The slope of the fitting curve between the predicted soil shear strength index c value and the actual value constructed by the random forest regression model using the method of this invention is 0.9624, and the correlation coefficient R reaches 0.986. The slope of the fitted curve is 0.9887, and the correlation coefficient R reaches 0.967.

[0067] A second aspect of the present invention provides a system for determining soil shear strength parameters based on random forest load tests, the system comprising:

[0068] The data acquisition module is used to acquire load test PS curves and corresponding soil shear strength index test data from actual engineering data, as well as load test PS curves and corresponding soil shear strength indexes simulated by numerical simulation software.

[0069] The dataset construction module is used to extract the feature data p of all load test ps curves from the data acquisition module. cr s cr a and b, and establish a training sample set, where p cr and s cr denoted as the bearing plate load and bearing plate settlement value corresponding to the end point of the straight line segment of the ps curve, and a and b are the fitting coefficients of the curve segment after the straight line segment of the ps curve.

[0070] The data training module trains a random forest regression model using the sample set in the dataset construction module, and uses the load test feature data p from the training sample set. cr s cr a and b are used as model inputs, and soil shear strength index c or For the output, train a random forest regression model, where c is the cohesion. It is the internal friction angle;

[0071] The practical engineering application module acquires load test ps curve data of the tested soil in actual engineering projects and extracts characteristic data p representing the load test ps curve of the tested soil in actual engineering projects. cr s cr a and b, and input the feature data into the random forest regression model obtained by the data training module, outputting the shear strength index c and b of the soil tested in the actual engineering load test. The value.

[0072] The system's program module provided by this invention can extract characteristic data for characterizing the entire process of the ps curve in a load test; then, using data obtained from actual engineering surveys and numerical simulations of load tests, it establishes a random forest regression analysis sample set for determining soil shear strength through load tests; and finally, it constructs a system with load test characteristic data as input and soil shear strength index c or... The output is a random forest regression analysis model, which is then learned and validated using sample set data to form a random forest regression model that determines the soil shear strength index from load test data. Finally, in engineering practice, feature data can be extracted from the load test data and input into the random forest regression model to obtain the soil shear strength index c or...

[0073] An embodiment of the third aspect of the present invention provides a computer-readable storage medium having an executable program stored thereon, wherein the executable program, when executed by a processor, implements the steps of the method for determining soil shear strength index based on random forest load tests as described above.

[0074] The processor-executable program of the computer-readable storage medium provided by this invention can extract characteristic data for characterizing the entire process of the ps curve of a load test; then, using data obtained from actual engineering surveys and numerical simulations of load tests, it establishes a random forest regression analysis sample set for determining soil shear strength through load tests; and then constructs a system with load test characteristic data as input and soil shear strength index c or... The output is a random forest regression analysis model, which is then learned and validated using sample set data to form a random forest regression model that determines the soil shear strength index from load test data. Finally, in engineering practice, feature data can be extracted from the load test data and input into the random forest regression model to obtain the soil shear strength index c or...

[0075] An embodiment of the fourth aspect of the present invention provides an apparatus for determining soil shear strength indices. The apparatus includes at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the steps of the above-described method for determining soil shear strength indices using a random forest-based load test.

[0076] The processor execution program of the device for determining soil shear strength index provided by this invention can extract characteristic data for characterizing the entire process of the load test ps curve; then, using data obtained from actual engineering surveys and load test numerical simulations, it establishes a random forest regression analysis sample set for determining soil shear strength by load test; and then constructs a system with load test characteristic data as input and soil shear strength index c or The output is a random forest regression analysis model, which is then learned and validated using sample set data to form a random forest regression model that determines the soil shear strength index from load test data. Finally, in engineering practice, feature data can be extracted from the load test data and input into the random forest regression model to obtain the soil shear strength index c or...

[0077] To better understand the method for determining soil shear strength indices according to the present invention, the method for determining soil shear strength indices based on random forest load tests according to the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0078] Example 1

[0079] This embodiment provides a method for extracting characteristic data from the ps curve of a load test.

[0080] Characteristic data representing the ps curve of the load test include the initial straight line segment p and the endpoint p. cr s cr The hyperbola fitting parameters a and b for the curve segment after the straight line segment of ps are given, wherein the endpoint of the initial straight line segment of ps is determined by the change in the slope of the ps curve being less than 0.5.

[0081] The specific change in the slope of the PS curve is determined by linear interpolation of the slope changes between two adjacent test points on the PS curve. The formula for calculating the change in the slope of a certain test point i on the PS curve is:

[0082]

[0083] in: s i p i These are the values ​​of s and p at measurement point i; s max p max These are the s and p values ​​corresponding to the last load level in this set of ps data.

[0084] The load p on the bearing plate corresponding to the endpoint of the straight segment of the ps curve is... cr The settlement value s of the bearing plate corresponding to the endpoint of the straight line segment of the ps curve. cr .

[0085] The fitting coefficient for the curve segment after the straight line segment of the PS curve is obtained by fitting the data points of the three load levels before the last load level in the load test onto a hyperbola. The fitting formula is as follows:

[0086] s / p = a·s + b,

[0087] Where: p is the load applied to the bearing plate, and s is the settlement under the corresponding load.

[0088] The following example uses measured load test data from a certain engineering project to illustrate the point:

[0089] (1) Actual engineering load test ps (load-settlement) test data are shown in Table 1, and the corresponding ps curves are as follows: Figure 2 As shown by the solid line in the middle.

[0090] Table 1. Results of site load test for a certain project.

[0091]

[0092] (2) The feature data representing the ps curve are extracted as shown in Table 2. The curve is backtracked using the feature values. Figure 2As shown by the dashed line, the straight line segment directly connects the origin of the coordinate system to (p). cr s cr Point; straight line segment followed by curve segment: based on s / p = a·s + b, use the extracted feature values ​​to solve for curve segment p. i The corresponding s i ', make backtracking p i -s i 'curve.

[0093] Table 2. Characteristic values ​​of ps curves from loading experiments used in random forest regression analysis.

[0094] Characteristic parameter p cr ]]> s cr ]]> a b Characteristic value 82.588 0.5833 0.00672612 0.00235347

[0095] like Figure 2 As shown in Table 2, the ps curve obtained by backtracking the extracted feature data (as shown in Table 2) matches the actual measured ps curve very well, proving that the initial straight line segment endpoint p of the extracted feature data ps is consistent with the actual measured ps curve. cr s cr The correlation height between the hyperbola fitting parameters a and b of the straight line segment and the curve segment of ps and the ps curve can truly reflect the shape of the ps curve.

[0096] Example 2

[0097] This embodiment provides a stochastic forest regression model for constructing soil shear strength indices based on load test data.

[0098] like Figure 3 The flowchart shown below illustrates the formation process of the random forest regression model. The specific steps are as follows:

[0099] Step S1: Simulate the ps curve using actual engineering surveys and load tests, and preprocess the ps curve data;

[0100] Step S2: Extract the feature data p of the ps curve from step S1. cr s cr a and b, and establish a training sample set;

[0101] Step S3: Train a random forest regression model using the training sample set, and use the load test feature data p from the training sample set. cr s cr a and b are used as model inputs, and the soil shear strength index c or For output, train a random forest regression model;

[0102] This embodiment uses a total of 2000 sets of load test data and related soil shear strength data, including 1950 sets obtained from numerical simulation and 50 sets of actual engineering data. Each set of data includes p cr s cra, b, c and The training set contains 1600 datasets, and the test set contains 400 datasets. Among these, the datasets include data from actual engineering projects. It refers to the shear strength test index of the soil corresponding to the load test; c in numerical simulation, These are the soil shear strength parameters given during the simulation.

[0103] Table 3 shows a partial sample set of data required to establish soil shear strength parameters based on random forest regression.

[0104] Table 3 shows partial sample data used for random forest regression analysis.

[0105]

[0106] In step S3, a soil shear strength index c is constructed using load test characteristic data as input or... The parameters of the random forest regression model can be set as shown in Table 4.

[0107] Table 4. Parameter values ​​for the random forest regression analysis model.

[0108]

[0109] Then, unsupervised learning training is performed using the sample set data to form a random forest regression model that determines the soil shear strength index based on load test data.

[0110] The evaluation results of the random forest regression model are as follows: Figure 4 and Figure 5 As shown in the figure, the horizontal axis represents the shear strength index c of the soil. The shear strength test values ​​from actual engineering data or the index values ​​used in numerical simulations are represented on the ordinate, which is the shear strength index value c obtained from the load test ps curve using a trained random forest regression model. From the fitted trend line equation and its correlation coefficient R of the data points in the graph 2 It can be seen that the method provided by the present invention has good analytical effect.

[0111] To better illustrate how this invention utilizes load test results to provide more accurate in-situ soil shear strength indices for engineering design, the applicant uses the method of this invention to determine the soil shear strength indices for the project based on actual load test data from real-world engineering projects.

[0112] Example 3

[0113] This embodiment uses the measured load test data of an engineering project in Embodiment 1 as an example, and applies the random forest regression model constructed in Embodiment 2 to solve for the soil shear strength index. The specific process is as follows:

[0114] Step S1: Obtain the actual engineering load test data (ps, load-settlement), as shown in Table 1, and the corresponding ps curve is shown below. Figure 2 As shown by the solid line in the middle.

[0115] Step S2: Extract the characteristic data representing the ps curve of the actual engineering load test, as shown in Table 2. Use the characteristic values ​​to backtrack the curve, as shown in Table 2. Figure 2 As shown by the dashed line;

[0116] Step S3: Input the extracted feature data into the trained random forest regression model that determines the soil shear strength index from the load test data, to obtain the shear strength index c of the soil tested in the load test and The numerical value is given; in this embodiment, the shear strength index c = 10 kPa.

[0117] Where there is no conflict, the above embodiments and features described herein can be combined with each other.

[0118] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for determining soil shear strength parameters based on random forest load tests, characterized in that, The specific steps include the following: S1. Load testing to obtain actual engineering data and software simulations. p - s Curve and corresponding soil shear strength index; S2. Extract all load tests from step S1. p - s Feature data of curves p cr , s cr , a and b And establish a training sample set, in which, p cr and s cr for p - s The load on the bearing plate and the settlement value of the bearing plate at the end of the straight section of the curve. a and b for p - s The fitting coefficient of the curve segment after the straight line segment; S3. Train a random forest regression model using the training sample set, and use the load test feature data from the training sample set. p cr , s cr , a and b As input to the model, the corresponding soil shear strength index c or φ For the output, train a random forest regression model, where c For cohesion, φ It is the internal friction angle; S4. Obtain load test results for the soil being tested in actual engineering projects. p - s Curve test data are used to extract load test data that characterizes the soil being tested in actual engineering projects. p - s Feature data of curves p cr , s cr , a and b The feature data is then input into the trained random forest regression model, which ultimately outputs the shear strength index of the soil tested under actual engineering load. c and φ The value; In step S2, the aforementioned p - s The endpoint of the straight segment of the curve passes through the... p - s The slope of the curve is determined to be less than 0.

5. The feature data a and b Data points from the three load levels preceding the last load level in the load test were fitted using a hyperbola. The fitting formula is as follows: s / p = a · s + b ， in, p For the load applied to the bearing plate, s This represents the settlement under the corresponding load.

2. The method as described in claim 1, characterized in that, The p - s The change in the slope of the curve is used p - s The slope change values ​​of two adjacent test points on the curve are determined by linear interpolation. p - s On a certain curve i The formula for calculating the change in slope at the test point is: ， in, , , s i , p i yes i measuring point s and p value; s max , p max yes p - s The last load level in the curve corresponds to s and p Value, of which, p For the load applied to the bearing plate, s This represents the settlement under the corresponding load.

3. The method as described in claim 1, characterized in that, In step S3, the random forest regression model is trained using the training sample set as follows: 80% of the training samples are randomly selected to train the random forest regression model, and the remaining 20% ​​of the samples are used to test the model.

4. The method as described in claim 1, characterized in that, In step S3, the parameters for setting the random forest regression model include: the maximum depth of each decision tree is set to 5, the number of random forest regression trees is set to 100, the minimum sample size for node splitting is 2, and the minimum sample size for leaf nodes is 1.

5. A system for determining soil shear strength parameters based on random forest load tests, characterized in that, The system includes: The data acquisition module is used to acquire load test data from actual engineering data. p - s Curve and corresponding soil Shear strength test data of the body and load test simulated using numerical simulation software. p - s Curve and corresponding soil shear strength index; The dataset construction module is used to extract all load experiments from the data acquisition module. p - s Feature data of curves p cr , s cr , a and b And establish a training sample set, in which, p cr and s cr for p - s The load on the bearing plate and the settlement value of the bearing plate at the end of the straight section of the curve. a and b for p - s The fitting coefficient of the curve segment after the straight line segment; The data training module trains a random forest regression model using the sample set in the dataset construction module, and incorporates load test feature data from the training sample set. p cr , s cr , a and b Soil shear strength index is used as model input. c or φ For the output, train a random forest regression model, where c For cohesion, φ It is the internal friction angle; The practical engineering application module acquires load test data of the soil being tested in actual engineering projects. p - s Curve test data are used to extract load test data that characterizes the soil being tested in actual engineering projects. p - s Feature data of curves p cr , s cr , a and b The feature data is then input into the random forest regression model trained by the data training module, and the shear strength index of the soil tested in the actual engineering load test is output. c and φ The value; The aforementioned p - s The endpoint of the straight segment of the curve passes through the... p - s The slope of the curve is determined to be less than 0.

5. The feature data a and b Data points from the three load levels preceding the last load level in the load test were fitted using a hyperbola. The fitting formula is as follows: s / p = a · s + b ， in, p For the load applied to the bearing plate, s This represents the settlement under the corresponding load.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which is configured to, when called by a processor, implement the steps of the method for determining soil shear strength index based on random forest load tests as described in any one of claims 1-4.

7. A device for determining the shear strength index of soil, characterized in that: The device includes at least one processor and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the method for determining soil shear strength index based on random forest load tests as described in any one of claims 1-4.

Images