Shallow foundation bearing capacity calculation method fusing machine learning and non-local algorithm

Abstract

The invention discloses a shallow foundation bearing capacity calculation method fusing machine learning and a non-local algorithm, which belongs to geotechnical engineering, and comprises the following steps: 1, judging foundation soil type and strain hardening type foundation soil according to an indoor triaxial test result, 2, carrying out 3 and 2, establishing a finite element model, carrying out refined simulation, and then carrying out 7, carrying out 4, carrying out 5, carrying out 6, carrying out 7, carrying out 7, carrying out 7, carrying out 7, carrying out 7, carrying out 7, carrying out 7, carrying out 7, carrying out 7, carrying out 7, and carrying out 7; 3, performing refined simulation by using a constitutive model combined with a non-local algorithm, and then turning to 4 and 4 to calibrate a shallow foundation analysis influence range DLf and a softening rate control parameter beta, 5, judging whether the DLf is appropriate, turning to 7 if the DLf is appropriate, and turning to 6 and 6 to construct and train a machine learning agent model if the DLf is not appropriate, and finally obtaining appropriate DL'f, beta 1 and 7 to perform shallow foundation analysis, and finally, performing the shallow foundation analysis. By the adoption of the method, efficient inversion of non-local parameters is achieved, and therefore the reasonable shallow foundation bearing capacity prediction value is obtained.

Description

Technical Field

[0001] The present invention relates to the field of geotechnical engineering technology, and in particular to a method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms. Background Art

[0002] Calculating the bearing capacity of shallow foundations during loading is crucial for geotechnical engineering design. Traditional standard methods for shallow foundation bearing capacity analysis are based on an elastic half-space theoretical model, which assumes a linear elastic relationship between soil stress and strain. This model deviates significantly from the nonlinear, elastoplastic, and stratified nature of actual soils. Furthermore, the determination of key parameters (such as the subgrade coefficient) in the standard relies primarily on empirical formulas or simplified assumptions, which can lead to calculations that deviate from actual engineering practice.

[0003] The calculation of bearing capacity in shallow foundations is an important part of geotechnical engineering. When using the finite element method to analyze shallow foundations built on a foundation with softening characteristics, a non-local algorithm is needed to eliminate the mesh dependency of the calculation results in the softening stage. In addition, the stress-strain characteristics of the soil are complex and require a reasonable constitutive model to describe them. However, when using a constitutive model combined with a non-local algorithm to analyze the bearing capacity of shallow foundations, the non-local algorithm should only change the softening rate of the load-displacement curve to eliminate the mesh dependency problem. However, the value of the non-local parameters has a significant impact on the residual state of the load-displacement curve, resulting in the inability to obtain a reasonable prediction value of the foundation bearing capacity.

[0004] Based on this, the present invention proposes a shallow foundation bearing capacity calculation method that integrates machine learning and non-local algorithms. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for calculating the bearing capacity of shallow foundations that integrates machine learning and non-local algorithms. Based on the simulation of indoor triaxial experiments, the influence range, softening rate and problem size of shallow foundation analysis are calibrated to calibrate the influence range and softening rate control parameters for shallow foundation analysis, and machine learning technology is used to enhance efficiency, thereby achieving efficient inversion of non-local parameters, thereby obtaining a reasonable prediction value of the shallow foundation bearing capacity.

[0006] To achieve the above objectives, the present invention provides a method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms, comprising the following steps:

[0007] S1. Determine the type of foundation soil based on the results of the indoor triaxial test. If the foundation soil type is strain hardening type, proceed to S2; if the foundation soil type is strain softening type, proceed to S3;

[0008] S2. According to the triaxial test conditions, a finite element model is established, and based on the selected constitutive model, the entire triaxial test process is simulated in detail, and the constitutive model parameters are calibrated, and then S7 is carried out;

[0009] S3. According to the triaxial test conditions, a finite element model is established, and a constitutive model combined with a non-local algorithm is used to perform a detailed simulation of the entire triaxial test process, and then proceed to S4;

[0010] S4. Calibration of constitutive model parameters and triaxial analysis influence range DL t and softening rate control parameter β, then the triaxial analysis influence range DL t Convert to shallow foundation analysis influence range DL f ;

[0011] S5. Determine the impact range of shallow foundation analysis DL based on the grid size used in shallow foundation analysis f Is it suitable? If so, transfer to S7; if not, transfer to S6;

[0012] S6. Build and train a machine learning agent model, and use the machine learning agent model to calibrate the constitutive model parameters and the influence range of the new triaxial test simulation and the new softening rate control parameters, and finally obtain the appropriate DL f ;

[0013] S7. Use the calibrated constitutive model parameters and appropriate shallow foundation analysis influence range and softening rate control parameters to perform shallow foundation analysis and obtain the predicted value of foundation bearing capacity.

[0014] Preferably, the process of determining the foundation soil type by indoor triaxial test results in S1 is as follows:

[0015] S11. Sample the selected foundation soil to obtain a sample foundation soil, and perform water injection, back pressure saturation, K0 consolidation, and pressure shear operations on the sample foundation soil according to the Geotechnical Test Method Standard GB / T50123-2019;

[0016] S12. During the pressure-shear process, record the state of the sample foundation soil every 1-2 seconds to obtain a stress-strain curve. Observe the stress-strain curve. If the shape of the curve increases with the increase of strain, the stress continues to increase and eventually reaches a peak value and no longer changes, then the foundation soil is a strain-hardening type of foundation soil, and proceed to S2. If the shape of the curve increases with the increase of strain, the stress first reaches a peak value and then begins to decrease and finally no longer changes, then the foundation soil is a strain-softening type of foundation soil, and proceed to S3.

[0017] Preferably, the process of S2 is as follows:

[0018] S21. Based on the actual size of the foundation soil sample used in the triaxial test in S1, a three-dimensional finite element model is established that is consistent with the sample size and takes fluid-solid coupling into consideration;

[0019] S22. Add contact surface elements to both ends of the established three-dimensional finite element model to simulate the end constraint effect of the test instrument on both ends of the specimen;

[0020] S23, applying a total stress increment taking into account the gravity gradient to the cylindrical surface of the sample foundation soil, and applying a pore pressure increment condition of zero to both ends of the sample, then applying a total stress increment to the cylindrical surface and top of the sample foundation soil, and applying a pore pressure increment depending on the in-situ pore water pressure of the sample foundation soil to both ends of the sample;

[0021] S24, applying different total stress increments to the cylindrical surface and both ends of the sample foundation soil, restoring the K0 consolidation state of the sample foundation soil, and setting the pore pressure increments at both ends of the sample to zero. Then, all pore water pressure degrees of freedom at the top and bottom ends of the sample are bound, applying a total stress increment of zero to the cylindrical surface, and applying a constant displacement increment to the top end for loading;

[0022] S25. Complete the refined simulation and obtain the stress-strain curve, which is then compared with the stress-strain test data measured in the triaxial test to obtain the calibrated constitutive model parameters, and then proceed to S7.

[0023] Preferably, S3 uses a constitutive model combined with a non-local algorithm to perform a refined simulation of the entire triaxial test process as follows:

[0024] S31. Based on the actual size of the foundation soil sample subjected to the triaxial test in S1, a three-dimensional finite element model is established that is consistent with the sample size and takes fluid-solid coupling into consideration;

[0025] S32. Adding contact surface elements to both ends of the established three-dimensional finite element model to simulate the end constraint effect of the test instrument on both ends of the specimen;

[0026] S33. Use the same method as S23-S24 to perform refined simulation to obtain the stress-strain curve, and then proceed to S4.

[0027] Preferably, the constitutive model parameters and the shallow foundation analysis influence range DL are calibrated in S4 f The process with the softening rate control parameter β is as follows:

[0028] S41. Select a suitable triaxial test influence range DL for triaxial test simulation t , DL t The value of is 1-2 times the mesh size used in the triaxial test simulation;

[0029] S42, comparing the stress-strain curve obtained in S33 with the stress-strain test data measured in the triaxial test, and calibrating the constitutive model parameters and the softening rate control parameter β;

[0030] S43, use the problem size to calibrate the triaxial test influence range DL in the triaxial test simulation t Scaled to shallow foundation analysis impact range DL for analyzing shallow foundation problems f , as follows:

[0031] DL f / B f =DL t / B t ;

[0032] Where B represents the problem size, subscript f represents shallow foundation analysis, and subscript t represents triaxial test.

[0033] Preferably, in S5, the impact range DL of the shallow foundation analysis is determined f Is it appropriate? There must be enough Gauss points to calculate the nonlocal strain, so that the strain concentrated on the shear band can be discretized to the surrounding units, reducing the stress / strain concentration on the shear band, thereby eliminating the mesh dependence in the finite element calculation. Therefore, if DL f The value of is greater than the size of the largest grid on the shear band, then β and DL f For appropriate softening rate control parameters and shallow foundation analysis influence range, go to S7, otherwise go to S6.

[0034] Preferably, the process of constructing and training the machine learning agent model in S6 and obtaining a suitable shallow foundation analysis calibration impact range is as follows:

[0035] S61. Using nonlinear multi-scale linear compression technology, the data characteristics of high-dimensional output space are proposed, and the output result set of finite element is converted into From the original space D Y Mapping to the reduced space D Z ;

[0036] S62, based on the reduced space D Z , a polynomial chaos expansion technique is used to replace the finite element model used in the above triaxial test simulation with a polynomial approximate function to construct a proxy model;

[0037] S63. Calculate the triaxial test simulation results using finite element method, use nonlinear multi-scale linear compression technology to extract data features of the output results, and train the agent model;

[0038] S64. Select a new triaxial test influence range DL' tand a new softening rate control parameter β1, and adjust the constitutive model parameters, and then use the surrogate model to output the triaxial test simulation results;

[0039] S65. If the simulation results obtained in S64 are inconsistent with the experimental results, return to S64 to readjust the constitutive model parameters and softening rate control parameters until the triaxial test simulation results match the experimental data. At this time, the calibrated constitutive model parameters and the new triaxial test influence range DL' are obtained. t and a new softening rate control parameter β1;

[0040] S66, DL' obtained in S65 t The new shallow foundation analysis influence range DL' is obtained by scaling using the method in S43 f , then judge DL' f Is it suitable? If not, return to S64. If suitable, then β1 and DL' f To determine the appropriate softening rate control parameters and the impact range of shallow foundation analysis, proceed to S7.

[0041] Preferably, the process of performing shallow foundation analysis in S7 to obtain a predicted value of foundation bearing capacity is as follows:

[0042] S71. If the foundation soil is a strain hardening type, the constitutive model parameters from S2 are received. If the foundation soil is a strain softening type, the constitutive model parameters from S3-S6 and the appropriate softening rate control parameters and shallow foundation analysis influence range are received.

[0043] S72. Use displacement-controlled loading on the shallow foundation and calculate the reaction force of the foundation on the shallow foundation through finite element calculation;

[0044] S73. Repeat the process of S72 to obtain the load-displacement curve of the shallow foundation. When the load-displacement curve tends to be stable, apply vertical displacement to the shallow foundation. Stop the finite element analysis when the reaction force of the foundation soil on the shallow foundation remains basically unchanged. The reaction force value obtained at this time is the predicted value of the foundation bearing capacity.

[0045] Therefore, the present invention adopts a shallow foundation bearing capacity calculation method that integrates machine learning and non-local algorithms, which has the following advantages:

[0046] 1. Solved the problem of not using a reasonable method to calibrate nonlocal parameters and obtain reasonable foundation bearing capacity prediction values when using the constitutive model combined with the nonlocal algorithm to analyze shallow foundations built on foundations with softening characteristics;

[0047] 2. The method proposed in this invention is based on the analysis of the influence of non-local parameters. It has no impact on the previous triaxial test process and does not require any additional indoor or outdoor experiments on shallow foundations.

[0048] 3. The method proposed in the present invention fully utilizes numerical simulation technology and machine learning technology to calibrate parameters. It can be automatically completed by a high-performance server and does not require consumables. Therefore, the labor cost and economic cost are low and the repeatability is high.

[0049] 4. By using the relationship between the influence range, softening rate and problem size in shallow foundation analysis, the constitutive model parameters and non-local parameters are calibrated for shallow foundation analysis involving softening. Machine learning technology is used to enhance efficiency and achieve efficient inversion of non-local parameters, thereby obtaining reasonable shallow foundation bearing capacity prediction values, while also taking into account the shallow foundation bearing capacity prediction values of hardened foundation soils.

[0050] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of a method for calculating the bearing capacity of a shallow foundation by integrating machine learning and non-local algorithms according to the present invention;

[0052] Figure 2 This is a workflow diagram of a machine learning agent model for a shallow foundation bearing capacity calculation method that integrates machine learning and non-local algorithms in the present invention. DETAILED DESCRIPTION

[0053] Example

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0055] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort shall fall within the scope of protection of the present invention.

[0056] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings.

[0057] In the description of the present invention, it should be noted that the terms "upper", "lower", "inside", "outside", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the inventive product is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they should not be understood as limiting the present invention.

[0058] In the description of the present invention, it should also be noted that, unless otherwise expressly specified or limited, the terms "disposed," "installed," and "connected" should be understood broadly. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.

[0059] The following embodiments of the present invention are described in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments may be combined with each other.

[0060] like Figure 1 and Figure 2 As shown, the present invention provides a method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms, comprising the following steps:

[0061] S1. Determine the type of foundation soil based on the results of the indoor triaxial test. If the foundation soil type is strain hardening type, proceed to S2; if the foundation soil type is strain softening type, proceed to S3;

[0062] S11. Sample the selected foundation soil to obtain a sample foundation soil, and perform water injection, back pressure saturation, K0 consolidation, and pressure shear operations on the sample foundation soil according to the Geotechnical Test Method Standard GB / T50123-2019;

[0063] S12. During the pressure-shear process, record the state of the sample foundation soil every 1-2 seconds to obtain a stress-strain curve. Observe the stress-strain curve. If the shape of the curve increases with the increase of strain, the stress continues to increase and eventually reaches a peak value and no longer changes, then the foundation soil is a strain-hardening type of foundation soil, and proceed to S2. If the shape of the curve increases with the increase of strain, the stress first reaches a peak value and then begins to decrease and finally no longer changes, then the foundation soil is a strain-softening type of foundation soil, and proceed to S3.

[0064] S2. According to the triaxial test conditions, a finite element model is established, and based on the selected constitutive model, the entire triaxial test process is simulated in detail, and the constitutive model parameters are calibrated, and then S7 is carried out;

[0065] S21. Based on the actual size of the foundation soil sample used in the triaxial test in S1, a three-dimensional finite element model is established that is consistent with the sample size and takes fluid-solid coupling into consideration;

[0066] S22. Add contact surface elements to both ends of the established three-dimensional finite element model to simulate the end constraint effect of the test instrument on both ends of the specimen;

[0067] S23, applying a total stress increment taking into account the gravity gradient to the cylindrical surface of the sample foundation soil, and applying a pore pressure increment of zero to both ends of the sample. Then, applying a total stress increment to the cylindrical surface and top of the sample foundation soil, and applying a pore pressure increment depending on the in-situ pore water pressure of the sample foundation soil to both ends of the sample. This step needs to be long enough to meet drainage conditions.

[0068] S24. Apply different total stress increments to the cylindrical surface and both ends of the sample foundation soil, restore the sample foundation soil to the K0 consolidation state, and set the pore pressure increments at both ends of the sample to zero. Then, bind all pore water pressure degrees of freedom at the top and bottom ends of the sample, apply a total stress increment of zero to the cylindrical surface, and apply a constant displacement increment to the top end for loading. The loading rate needs to be determined according to the triaxial test conditions.

[0069] S25. Complete the refined simulation and obtain the stress-strain curve, which is then compared with the stress-strain test data measured in the triaxial test to obtain the calibrated constitutive model parameters, and then proceed to S7.

[0070] S3. According to the triaxial test conditions, a finite element model is established, and a constitutive model combined with a non-local algorithm is used to perform a detailed simulation of the entire triaxial test process, and then proceed to S4;

[0071] S31. Based on the actual size of the foundation soil sample subjected to the triaxial test in S1, a three-dimensional finite element model is established that is consistent with the sample size and takes fluid-solid coupling into consideration;

[0072] S32. Adding contact surface elements to both ends of the established three-dimensional finite element model to simulate the end constraint effect of the test instrument on both ends of the specimen;

[0073] S33. Use the same method as S23-S24 to perform refined simulation to obtain the stress-strain curve, and then proceed to S4.

[0074] S4. Calibration of constitutive model parameters and triaxial analysis influence range DL t and softening rate control parameter β, then the triaxial analysis influence range DL t Convert to shallow foundation analysis influence range DL f ;

[0075] S41. Select a suitable triaxial test influence range DL for triaxial test simulation t , DL t The value of is 1-2 times the mesh size used in the triaxial test simulation;

[0076] S42, comparing the stress-strain curve obtained in S33 with the stress-strain test data measured in the triaxial test, and calibrating the constitutive model parameters and the softening rate control parameter β;

[0077] S43, use the problem size to calibrate the triaxial test influence range DL in the triaxial test simulation t Scaled to shallow foundation analysis impact range DL for analyzing shallow foundation problems f , as follows:

[0078] DL f / B f =DL t / B t ;

[0079] Where B represents the problem size, subscript f represents shallow foundation analysis, and subscript t represents triaxial test.

[0080] S5. Determine the impact range of shallow foundation analysis DL based on the grid size used in shallow foundation analysis f Is it appropriate? If it is appropriate, go to S7; if not, go to S6. DL is the parameter that controls the influence range in the non-local algorithm, that is, all Gaussian points within the radius of DL around the current Gaussian point will participate in the calculation of non-local strain.

[0081] Determine the impact range of shallow foundation analysis DL f Is it appropriate? There must be enough Gauss points to calculate the nonlocal strain, so that the strain concentrated on the shear band can be discretized to the surrounding units, reducing the stress / strain concentration on the shear band, thereby eliminating the mesh dependence in the finite element calculation. Therefore, if DL f The value of is greater than the size of the largest grid on the shear band, then β and DL f For appropriate softening rate control parameters and shallow foundation analysis influence range, go to S7, otherwise go to S6.

[0082] S6. Build and train a machine learning agent model, and use the machine learning agent model to calibrate the constitutive model parameters and the influence range of the new triaxial test simulation and the new softening rate control parameters, and finally obtain the appropriate DL f ;

[0083] S61, using nonlinear multi-scale linear compression technology (MDS), proposes the data characteristics of high-dimensional output space, and sets the output result of finite element From the original space DY Mapping to the reduced space D Z ;

[0084] S62, based on the reduced space D Z , the polynomial chaos expansion (PCE) technique was used to replace the finite element model used in the above triaxial test simulation with a polynomial approximation function to construct a proxy model;

[0085] S63. Calculate the triaxial test simulation results using finite element method, use nonlinear multi-scale linear compression technology to extract data features of the output results, and train the agent model;

[0086] S64. Select a new triaxial test influence range DL' t and a new softening rate control parameter β1, and adjust the constitutive model parameters, and then use the surrogate model to output the triaxial test simulation results;

[0087] S65. If the simulation results obtained in S64 are inconsistent with the experimental results, return to S64 to readjust the constitutive model parameters and softening rate control parameters until the triaxial test simulation results match the experimental data. At this time, the calibrated constitutive model parameters and the new triaxial test influence range DL' are obtained. t and a new softening rate control parameter β1;

[0088] S66, DL' obtained in S65 t The new shallow foundation analysis influence range DL' is obtained by scaling using the method in S43 f , then judge DL' f Is it suitable? If not, return to S64. If suitable, then β1 and DL' f To determine the appropriate softening rate control parameters and the impact range of shallow foundation analysis, proceed to S7.

[0089] S7. Use the calibrated constitutive model parameters and appropriate shallow foundation analysis influence range and softening rate control parameters to perform shallow foundation analysis and obtain the predicted value of foundation bearing capacity.

[0090] S71. If the foundation soil is a strain hardening type, the constitutive model parameters from S2 are received. If the foundation soil is a strain softening type, the constitutive model parameters from S3-S6 and the appropriate softening rate control parameters and shallow foundation analysis influence range are received.

[0091] S72. Use displacement-controlled loading on the shallow foundation and calculate the reaction force of the foundation on the shallow foundation through finite element calculation;

[0092] S73. Repeat the process of S72 to obtain the load-displacement curve of the shallow foundation. When the load-displacement curve tends to be stable, apply vertical displacement to the shallow foundation. Stop the finite element analysis when the reaction force of the foundation soil on the shallow foundation remains basically unchanged. The reaction force value obtained at this time is the predicted value of the foundation bearing capacity.

[0093] Therefore, the present invention adopts a shallow foundation bearing capacity calculation method that integrates machine learning and non-local algorithms with the above structure, which solves the problem of not using a reasonable method to calibrate non-local parameters and obtain a reasonable foundation bearing capacity prediction value when using a constitutive model analysis combined with a non-local algorithm to analyze shallow foundations built on foundations with softening characteristics. Based on the analysis of the influence rules of non-local parameters, it has no impact on the previous triaxial test process, and there is no need to conduct any additional indoor or outdoor experiments on shallow foundations. It also uses numerical simulation technology and machine learning technology to calibrate parameters, which can be automatically completed by a high-performance server without consumables. Therefore, the labor cost and economic cost are low and the repeatability is high. At the same time, the relationship between the influence range, softening rate and problem size of shallow foundation analysis is used to calibrate the constitutive model parameters and non-local parameters for shallow foundation analysis involving softening. Machine learning technology is used to enhance efficiency, and efficient inversion of non-local parameters is achieved, thereby obtaining a reasonable shallow foundation bearing capacity prediction value, while taking into account the shallow foundation bearing capacity prediction value of hardened foundation soil.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A shallow foundation bearing capacity calculation method integrating machine learning and non-local algorithm, characterized in that: The following steps are involved: S1. Determine the type of foundation soil based on the indoor triaxial test results. If the foundation soil type is strain hardening type, proceed to S2; if the foundation soil type is strain softening type, proceed to S3; S2. According to the triaxial test conditions, a finite element model is established, and based on the selected constitutive model, the entire triaxial test process is simulated in detail, and the constitutive model parameters are calibrated, and then S7 is carried out; S3. According to the triaxial test conditions, a finite element model is established, and a constitutive model combined with a non-local algorithm is used to perform a detailed simulation of the entire triaxial test process, and then proceed to S4; S4. Calibration of constitutive model parameters and triaxial analysis influence range DL t and softening rate control parameter β, then the triaxial analysis influence range DL t Convert to shallow foundation analysis influence range DL f ; S5. Determine the impact range of shallow foundation analysis DL based on the grid size used in shallow foundation analysis f Is it suitable? If so, transfer to S7; if not, transfer to S6; S6. Build and train a machine learning agent model, and use the machine learning agent model to calibrate the constitutive model parameters and the influence range of the new triaxial test simulation and the new softening rate control parameters, and finally obtain the appropriate DL f ; S7. Use the calibrated constitutive model parameters and appropriate shallow foundation analysis influence range and softening rate control parameters to perform shallow foundation analysis and obtain the predicted value of foundation bearing capacity.

2. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 1 is characterized in that: The process of determining the foundation soil type based on the indoor triaxial test results in S1 is as follows: S11. Sample the selected foundation soil to obtain a sample foundation soil, and perform water injection, back pressure saturation, K0 consolidation, and pressure shear operations on the sample foundation soil according to the Geotechnical Test Method Standard GB / T 50123-2019; S12. During the pressure-shear process, record the state of the sample foundation soil every 1-2 seconds to obtain a stress-strain curve. Observe the stress-strain curve. If the shape of the curve increases with the increase of strain, the stress continues to increase and eventually reaches a peak value and no longer changes, then the foundation soil is a strain-hardening type of foundation soil, and proceed to S2. If the shape of the curve increases with the increase of strain, the stress first reaches a peak value and then begins to decrease and finally no longer changes, then the foundation soil is a strain-softening type of foundation soil, and proceed to S3.

3. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 2 is characterized in that: The process of S2 is as follows: S21. Based on the actual size of the foundation soil sample used in the triaxial test in S1, a three-dimensional finite element model is established that is consistent with the sample size and takes fluid-solid coupling into consideration; S22. Add contact surface elements to both ends of the established three-dimensional finite element model to simulate the end constraint effect of the test instrument on both ends of the specimen; S23, applying a total stress increment taking into account the gravity gradient to the cylindrical surface of the sample foundation soil, and applying a pore pressure increment condition of zero to both ends of the sample, then applying a total stress increment to the cylindrical surface and top of the sample foundation soil, and applying a pore pressure increment depending on the in-situ pore water pressure of the sample foundation soil to both ends of the sample; S24, applying different total stress increments to the cylindrical surface and both ends of the sample foundation soil, restoring the K0 consolidation state of the sample foundation soil, and setting the pore pressure increments at both ends of the sample to zero. Then, all pore water pressure degrees of freedom at the top and bottom ends of the sample are bound, applying a total stress increment of zero to the cylindrical surface, and applying a constant displacement increment to the top end for loading; S25. Complete the refined simulation and obtain the stress-strain curve, which is then compared with the stress-strain test data measured in the triaxial test to obtain the calibrated constitutive model parameters, and then proceed to S7.

4. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 3 is characterized in that: S3 uses a constitutive model combined with a non-local algorithm to perform a detailed simulation of the entire triaxial test process as follows: S31. Based on the actual size of the foundation soil sample subjected to the triaxial test in S1, a three-dimensional finite element model is established that is consistent with the sample size and takes fluid-solid coupling into consideration; S32. Adding contact surface elements to both ends of the established three-dimensional finite element model to simulate the end constraint effect of the test instrument on both ends of the specimen; S33. Use the same method as S23-S24 to perform refined simulation to obtain the stress-strain curve, and then proceed to S4.

5. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 4 is characterized in that: Calibration of constitutive model parameters and shallow foundation analysis influence range DL in S4 f The process with the softening rate control parameter β is as follows: S41. Select a suitable triaxial test influence range DL for triaxial test simulation t , DL t The value of is 1-2 times the mesh size used in the triaxial test simulation; S42, comparing the stress-strain curve obtained in S33 with the stress-strain test data measured in the triaxial test, and calibrating the constitutive model parameters and the softening rate control parameter β; S43, use the problem size to calibrate the triaxial test influence range DL in the triaxial test simulation t Scaled to shallow foundation analysis impact range DL for analyzing shallow foundation problems f , as follows: DL f / B f =DL t / B t ; Where B represents the problem size, subscript f represents shallow foundation analysis, and subscript t represents triaxial test.

6. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 5, characterized in that: Determine the impact range of shallow foundation analysis in S5 DL f Is it appropriate? There must be enough Gauss points to calculate the nonlocal strain, so that the strain concentrated on the shear band can be discretized to the surrounding units, reducing the stress / strain concentration on the shear band, thereby eliminating the mesh dependence in the finite element calculation. Therefore, if DL f The value of is greater than the size of the largest grid on the shear band, then β and DL f For appropriate softening rate control parameters and shallow foundation analysis influence range, go to S7, otherwise go to S6.

7. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 6 is characterized in that: The process of building and training the machine learning agent model in S6 and obtaining the appropriate shallow foundation analysis calibration impact range is as follows: S61. Using nonlinear multi-scale linear compression technology, the data characteristics of high-dimensional output space are proposed, and the output result set of finite element is converted into From the original space D Y Mapping to the reduced space D Z ; S62, based on the reduced space D Z , a polynomial chaos expansion technique is used to replace the finite element model used in the above triaxial test simulation with a polynomial approximate function to construct a proxy model; S63. Calculate the triaxial test simulation results using finite element method, use nonlinear multi-scale linear compression technology to extract data features of the output results, and train the agent model; S64. Select a new triaxial test influence range DL' t and a new softening rate control parameter β1, and adjust the constitutive model parameters, and then use the surrogate model to output the triaxial test simulation results; S65. If the simulation results obtained in S64 are inconsistent with the experimental results, return to S64 to readjust the constitutive model parameters and softening rate control parameters until the triaxial test simulation results match the experimental data. At this time, the calibrated constitutive model parameters and the new triaxial test influence range DL' are obtained. t and a new softening rate control parameter β1; S66, DL' obtained in S65 t The new shallow foundation analysis influence range DL' is obtained by scaling using the method in S43 f , then judge DL' f Is it suitable? If not, return to S64. If suitable, then β1 and DL' f To determine the appropriate softening rate control parameters and the impact range of shallow foundation analysis, proceed to S7.

8. The method for calculating the bearing capacity of shallow foundations by integrating machine learning and non-local algorithms according to claim 7 is characterized in that: The process of performing shallow foundation analysis in S7 and obtaining the predicted value of foundation bearing capacity is as follows: S71. If the foundation soil is a strain hardening type, the constitutive model parameters from S2 are received. If the foundation soil is a strain softening type, the constitutive model parameters from S3-S6 and the appropriate softening rate control parameters and shallow foundation analysis influence range are received. S72. Use displacement-controlled loading on the shallow foundation and calculate the reaction force of the foundation on the shallow foundation through finite element calculation; S73. Repeat the process of S72 to obtain the load-displacement curve of the shallow foundation. When the load-displacement curve tends to be stable, apply vertical displacement to the shallow foundation. Stop the finite element analysis when the reaction force of the foundation soil on the shallow foundation remains basically unchanged. The reaction force value obtained at this time is the predicted value of the foundation bearing capacity.

Images