Pipe pile bearing capacity prediction method based on igwo-bp neural network

Abstract

The application discloses a pipe pile bearing capacity prediction method based on an IGWO-BP neural network, and comprises the following steps: 1, pile-soil finite element models considering different pipe pile sizes and soil physical parameters are respectively established to perform dynamic load and static load calculation, and dynamic and static load test data are collected; 2, the dynamic load data obtained by calculation are subjected to feature selection by using a Pearson correlation coefficient; 3, a BP neural network model is trained by using the dynamic and static load test data, and an improved grey wolf optimization algorithm is used to optimize the weight values and threshold values between a hidden layer and an output layer in the BP neural network, and the best weight values and threshold values are found through global search; and 4, dynamic load test data of a prediction set are substituted into the IGWO-BP network model to perform pipe pile bearing capacity prediction. The application can use pipe pile dynamic load test data to predict corresponding pipe pile bearing capacity, so that high-precision pipe pile bearing capacity prediction can be realized, and problems such as high cost and long time consumption of pipe pile bearing capacity static load test can be overcome.

Description

Technical Field

[0001] This invention relates to the field of pipe pile bearing capacity testing, specifically a pipe pile bearing capacity prediction method based on an improved IGWO-BP neural network. Background Technology

[0002] Prestressed concrete pipe piles have advantages such as high single pile bearing capacity, flexible selection range, applicability to more geological and topographical conditions, controllable pile length, individual design according to different working conditions, less restriction on on-site construction, fast construction speed, and high growth rate. Compared with traditional cast-in-place piles of the same size, the single pile bearing capacity is increased by 30% to 50%. The main purpose of piles is to transfer and bear the load applied by the superstructure. The bearing capacity of piles mainly depends on two aspects: the quality of the pile itself and the mechanical properties of the foundation soil. If the pile has insufficient bearing capacity, it may cause uneven settlement, which may pose a major safety hazard to the later operation of the bridge. Therefore, the bearing capacity testing of this pile is particularly important. Pipe pile bearing capacity testing is mainly divided into static load method and dynamic load method. The static load method yields more accurate results, but it has disadvantages such as long time consumption and high cost. On the contrary, the dynamic load method is relatively cheaper and faster than the static load method, but the measured results have a larger error than the actual value, and it is currently difficult to use in actual engineering. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, this invention proposes a method for predicting the bearing capacity of pipe piles based on the IGWO-BP neural network, aiming to achieve high-precision prediction of the bearing capacity of pipe piles and overcome many problems such as high cost and long time consumption of static load tests for the bearing capacity of pipe piles.

[0004] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0005] The present invention provides a method for predicting the bearing capacity of pipe piles based on an IGWO-BP neural network, characterized by the following steps:

[0006] Step 1: Use finite element software to establish numerical models of the pipe pile and the soil respectively, and obtain the finite element model of the pipe pile and the finite element model of the soil, thus forming the pile-soil finite element model.

[0007] Step 2: Simulate the pile-soil behavior under real conditions by changing the pile size and soil physical parameters in the pile-soil finite element model;

[0008] Step 3: Obtain dynamic and static load test data, including: the bearing capacity characteristic vector of the pipe pile. and the displacement characteristic vector of the pile top ,in, These represent the bearing capacity characteristic data during the i-th static load test, This represents the displacement characteristic data during the i-th dynamic load test;

[0009] Step 4: Calculation Each displacement eigenvalue and The Pearson correlation coefficients between them were calculated, and all Pearson correlation coefficients were sorted in descending order, with the top [number] being selected. The displacement characteristic values ​​corresponding to each Pearson correlation coefficient are used to obtain the displacement characteristic data after the i-th dynamic load test screening. Thus, the filtered displacement feature vectors are obtained. ,in, express The first in Each displacement characteristic value;

[0010] Step 5: Create a BP neural network and use... As input to the BP neural network model, The output of the BP neural network model is used to train the BP neural network and obtain the trained pipe pile bearing capacity prediction model.

[0011] Step 6: Optimize the weights and thresholds of the trained pipe pile bearing capacity prediction model using the improved gray wolf optimization algorithm to obtain the optimal weights and thresholds;

[0012] Step 7: Use the trained pipe pile bearing capacity measurement model corresponding to the optimal weight and optimal threshold as the optimal pipe pile bearing capacity prediction model to realize the prediction of pipe pile bearing capacity.

[0013] The method for predicting the bearing capacity of pipe piles based on the IGWO-BP neural network described in this invention is characterized in that step 2 includes the following steps:

[0014] Step 2.1: Randomly vary the outer diameter, inner diameter, and length of the pipe pile in the finite element model within a certain range to simulate the dimensions of different pipe piles in actual engineering.

[0015] Step 2.2: Randomly vary the elastic modulus, cohesion, and internal friction angle of each soil layer in the finite element model within a certain range to simulate soil properties under different geological conditions;

[0016] Step 2.3: In the pile-soil finite element model, the friction coefficient between the pile and soil contact surfaces is randomly varied within a certain range to obtain data on the pile-soil contact behavior in actual engineering.

[0017] Furthermore, step 3 includes the following steps:

[0018] Step 3.1: By applying graded loads to the pile top, the load-settlement curve of the pile under graded loads is calculated. Based on the vertical load-settlement curve, the ultimate vertical compressive bearing capacity of a single pile is set as the load value corresponding to a total settlement of Δ. The characteristic value of the vertical compressive bearing capacity of a single pile is set as half of the ultimate vertical compressive bearing capacity of a single pile. A static load test is then performed on the pile-soil finite element model to obtain the characteristic vector of the pipe pile's bearing capacity. ,in, These represent the bearing capacity characteristic data at the i-th static load test; T represents transpose; n is the total number of tests;

[0019] Step 3.2: Create a finite element model with a heavy hammer to apply an impact load to the pile, thereby conducting a dynamic load test on the pile-soil finite element model and obtaining the displacement characteristic vector at the pile top. ,in, Let represent the displacement characteristic data during the i-th dynamic load test, and , express The j-th displacement characteristic value in the equation, where M is the number of displacement values.

[0020] Furthermore, step 6 includes the following steps:

[0021] Step 6.1: Initialize the maximum number of iterations Define the current iteration number as t, and initialize t=1;

[0022] The weights and thresholds of the trained pipe pile bearing capacity prediction model are used as the values ​​for the s-th gray wolf in the t-th generation. Thus, a wolf pack of size N is constructed in the tth generation.

[0023] Step 6.2: Calculate the value of the s-th gray wolf in the t-th generation wolf pack using equation (1). fitness value The algorithm obtains and sorts all gray wolf individuals in the t-th generation wolf pack, and then selects the three gray wolf individuals with the lowest fitness values ​​as the α wolf individuals in the t-th generation wolf pack. β wolf individuals δ wolf individual ;

[0024] (1)

[0025] In equation (1), The sth individual gray wolf The corresponding pipe pile bearing capacity prediction model The predicted value;

[0026] Step 6.3: Use the GWO algorithm to... The update is performed to obtain the s-th candidate gray wolf individual of the (t+1)-th generation. ;

[0027] Step 6.4: Use the DLH algorithm to... The update is performed, resulting in the s-th candidate gray wolf individual of generation t+1. ;

[0028] Step 6.5: Calculate using equation (1) and The fitness value is used to select the gray wolf individual with the smaller fitness value as the s-th gray wolf individual in the (t+1)-th generation. Thus, the (t+1)th generation wolf pack is obtained;

[0029] Step 6.6: After assigning t+1 to t, return to step 6.2 and execute sequentially until... Until then, thus obtaining the first We select N gray wolves from the wolf pack and choose the gray wolf with the lowest fitness as the optimal weight and optimal threshold for the pipe pile bearing capacity prediction model after training.

[0030] Furthermore, step 6.3 includes the following steps:

[0031] Step 6.3.1: Calculate the value of the s-th gray wolf using equation (2). Each with an α wolf individual Distance between β wolf individuals Distance between δ wolf individual Distance between :

[0032] (2)

[0033] In equation (2), , , Let represent the three first coefficient vectors, and we have:

[0034] (3)

[0035] In equation (3), , , Let represent three t-th generation random vectors within the range [0,1].

[0036] Step 6.3.2: Calculate and update the s-th gray wolf individual using equation (4). ;

[0037] (4)

[0038] In equation (4), , , These represent individuals based on α wolves. β wolf individuals δ wolf individual The sth individual gray wolf after the update; , , Let represent 3 second coefficient vectors, and we have:

[0039] (5)

[0040] In equation (5), , , Let represent the other 3 t-th generation random vectors in the range [0,1]. Let t represent the update vector for the t-th generation; and ;

[0041] Step 6.3.3: Use equation (6) to calculate the s-th candidate gray wolf individual in the (t+1)-th generation. ;

[0042] (6).

[0043] Furthermore, step 6.4 includes the following steps:

[0044] Step 6.4.1: Calculate using equation (7) and radius between :

[0045] (7)

[0046] In equation (7), Represents Euclidean distance;

[0047] Step 6.4.2: Calculate the s-th candidate gray wolf individual in the (t+1)-th generation using equation (8). ;

[0048] (8)

[0049] In equation (8); , respectively Centered on the radius Two randomly generated gray wolves; Let represent the random number in the t-th generation.

[0050] The present invention provides an electronic device, comprising a memory and a processor, wherein the memory is used to store a program that supports the processor in executing the pile bearing capacity prediction method, and the processor is configured to execute the program stored in the memory.

[0051] The present invention discloses a computer-readable storage medium on which a computer program is stored, wherein the computer program, when executed by a processor, performs the steps of the pipe pile bearing capacity prediction method.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0053] 1. Compared to static load testing, this invention does not require a huge investment of funds and time, and has advantages such as high efficiency and short cycle. It overcomes the disadvantages of static load testing, such as high cost, long time consumption, and terrain limitations, thus effectively solving many problems such as the long time consumption and high cost of predicting static load tests for pipe piles, and the difficulties that may be encountered in conducting static load tests for pipe piles under certain complex terrain or soil conditions.

[0054] 2. Compared with dynamic load testing, this invention achieves high-precision prediction of the bearing capacity of pipe piles by combining dynamic load test results with intelligent algorithms, thereby overcoming the problem of large errors in the results obtained from dynamic load testing.

[0055] 3. The model proposed in this invention has universality. By changing the width and height of the pipe pile, the elastic modulus of the soil and other physical parameters, it overcomes the problems of different pipe pile sizes and soil type variations, and thus can be applied to the prediction of pipe pile bearing capacity in different regions.

[0056] 4. This invention effectively finds the relationship between dynamic load displacement data and static load bearing capacity of pipe piles through a BP neural network. Thus, the change in bearing capacity can be predicted by the change in dynamic load displacement of pipe piles. By introducing an improved gray wolf optimization algorithm to optimize the initial weights and initial thresholds of the BP neural network, the prediction accuracy of the pipe pile bearing capacity prediction model is effectively improved. Attached Figure Description

[0057] Figure 1 This is a schematic diagram of the specific process of the method of the present invention;

[0058] Figure 2 This is a front view of the pile-hammer-soil model of the present invention;

[0059] Figure 3 This is a schematic diagram showing the dimensions and mesh generation of the finite element model of the present invention;

[0060] Figure 4 This is a flowchart of the IGWO algorithm of the present invention;

[0061] Figure 5This is a schematic diagram of the BP neural network structure of the present invention;

[0062] Figure 6 To develop a comparison chart of predicted and actual bearing capacity values ​​for pipe piles. Detailed Implementation

[0063] In this embodiment, refer to Figure 1 A method for predicting the bearing capacity of pipe piles based on IGWO-BP neural network includes the following steps:

[0064] Step 1: Use finite element software to establish numerical models of the pipe pile and the soil respectively, and obtain the finite element model of the pipe pile and the finite element model of the soil, thus forming the pile-soil finite element model.

[0065] In this embodiment, finite element models of the soil and the pipe pile are established using finite element software, referring to... Figure 3 part b in Figure 3 Part c in the model yields the pile-soil finite element model. The mesh generation of the pile-soil finite element model is based on... Figure 3 The d-section of the model is as follows: The soil is divided into 6 and 7 layers. With 6 layers, there are 3 contact layers between the pile and the soil; with 7 layers, there are 4 contact layers. To eliminate boundary effects, the soil width in the finite element model is generally 20 times the pile diameter, and the depth is approximately twice the pile penetration depth. Therefore, the soil is constructed as a rectangle with a length and width of 10m, a depth of 40m for 3-layer pile-soil contact, and 50m for 4-layer contact. With 3-layer pile-soil contact, the soil is divided into five layers with thicknesses of 5m, 5m, 12m, 9m, and 9m from top to bottom. With 4-layer pile-soil contact, the soil is divided into six layers with thicknesses of 5m, 5m, 5m, 15m, 10m, and 10m from top to bottom. The soil density is 2000... The parameters are: kg / m³, Poisson's ratio 0.3, elastic modulus 40 MPa, cohesion 20 kPa, and internal friction angle 20°. Later, by changing some of these parameters, different soil conditions under actual conditions are simulated. In the pile-soil finite element model, the friction coefficient at the pile-soil contact surface is taken as 0.35. Since the dynamic load test requires an additional hammer to apply the impact load, a separate hammer finite element model needs to be created when creating the finite element model. (Refer to...) Figure 3 Part a, shaped as a cylinder with a diameter of 1.2m and a height of 0.8m, has an elastic modulus of 200GPa and a weight of 6t. It is allowed to fall freely from a height of 70cm above the pile top. In the actual finite element modeling, the hammer is placed against the pile top and given an initial velocity of 3.74m / s to achieve the same effect as free fall. A frontal schematic diagram of the pile-hammer-soil model is shown below. Figure 2The finite element model of the pipe pile has an outer diameter of 0.8m, an inner diameter of 0.6m, a length of 18m when in contact with three layers of soil, a length of 23m when in contact with four layers of soil, a density of 2500 kg / m3, an elastic modulus of 48GPa, and a Poisson's ratio of 0.2.

[0066] Step 2: Simulate the pile-soil behavior under real conditions by changing the pile size and soil physical parameters in the pile-soil finite element model;

[0067] Step 2.1: Randomly vary the outer diameter, inner diameter, and length of the pipe pile in the finite element model within a certain range to simulate the dimensions of different pipe piles in actual engineering.

[0068] In this embodiment, the outer diameter and inner diameter of the pipe pile in the finite element model are multiplied by their respective random numbers. When in contact with the third and fourth layers of soil respectively, the pile length is multiplied by its respective random number. Finite element models of pipe piles of different sizes were obtained to simulate different pile sizes in real-world situations.

[0069] Step 2.2: Randomly vary the elastic modulus, cohesion, and internal friction angle of each soil layer in the finite element model within a certain range to simulate soil properties under different geological conditions;

[0070] In this embodiment, the elastic modulus, cohesion, and internal friction angle of the soil in the finite element model are multiplied by their respective random numbers. When the pile and soil are in contact at three layers, the thickness of the first, second, and fourth soil layers is multiplied by their respective random numbers. When the pile and soil are in contact at four layers, the thickness of the first, second, third, and fifth soil layers is multiplied by their respective random numbers. We obtained finite element models of soil with different parameters to simulate different soil conditions in real-world situations.

[0071] Step 2.3: In the pile-soil finite element model, the friction coefficient between the pile and soil contact surfaces is randomly varied within a certain range to obtain data on the pile-soil contact behavior in actual engineering.

[0072] In this embodiment, the friction coefficient of the pile-soil contact surface in the pile-soil finite element model is multiplied by a random number. Finite element models of piles and soil with different friction coefficients were obtained to simulate the pile-soil contact behavior under actual conditions.

[0073] Step 3: Obtain dynamic and static load test data:

[0074] By changing the pile-soil parameters, 200 sets of four-layer pile-soil contact models and 100 sets of three-layer pile-soil contact models were generated, and static load and dynamic load tests were performed respectively.

[0075] Step 3.1: By applying graded loads to the pile top, the load-settlement curve of the pile under graded loads is calculated. Based on the vertical load-settlement curve, the ultimate vertical compressive bearing capacity of a single pile is taken as the load value corresponding to a total settlement Δ = 40 mm, and the characteristic value of the vertical compressive bearing capacity of a single pile is taken as half of the ultimate vertical compressive bearing capacity of a single pile. The magnitudes of the first, second, and third graded loads are 1500 KN, 2000 KN, and 2400 KN, respectively, with the last graded load being 7000 KN. The intermediate load grades increase by 200 KN sequentially. A static load test is then performed on the pile-soil finite element model to obtain the characteristic vector of the pipe pile's bearing capacity. ,in, represents the bearing capacity characteristic data of the i-th static load test; T represents transpose; n is the total number of tests.

[0076] Step 3.2: Create a finite element model with a heavy hammer to apply an impact load to the pile, thereby conducting a dynamic load test on the pile-soil finite element model and obtaining the displacement characteristic vector at the pile top. ,in, Let represent the displacement characteristic data during the i-th dynamic load test, and , express The j-th displacement characteristic value in the equation, where M is the number of displacement values.

[0077] Step 4: Calculation Each displacement eigenvalue and The Pearson correlation coefficients between them were calculated, and all Pearson correlation coefficients were sorted in descending order, with the top [number] being selected. The displacement characteristic values ​​corresponding to each Pearson correlation coefficient are used to obtain the displacement characteristic data after the i-th dynamic load test screening. Thus, the filtered displacement feature vectors are obtained. ,in, express The first in Each displacement characteristic value.

[0078] Step 5: Refer to Figure 5 Create a BP neural network, including: an input layer, hidden layers, and an output layer; and use... As input to the BP neural network model, The output of the BP neural network model is used to train the BP neural network and obtain the trained pipe pile bearing capacity prediction model.

[0079] Step 6: Refer to Figure 4 The improved gray wolf optimization algorithm is used to optimize the weights and thresholds of the trained pipe pile bearing capacity prediction model, thereby obtaining the optimal weights and thresholds.

[0080] Step 6.1: Initialize the maximum number of iterations Define the current iteration number as t, and initialize t=1;

[0081] The weights and thresholds of the trained pipe pile bearing capacity prediction model are used as the values ​​for the s-th gray wolf in the t-th generation. Thus, a wolf pack of size N is constructed in the tth generation.

[0082] Step 6.2: Calculate the value of the s-th gray wolf in the t-th generation wolf pack using equation (1). fitness value The algorithm obtains and sorts all gray wolf individuals in the t-th generation wolf pack, and then selects the three gray wolf individuals with the lowest fitness values ​​as the α wolf individuals in the t-th generation wolf pack. β wolf individuals δ wolf individual ;

[0083] (1)

[0084] In equation (1), The sth individual gray wolf The corresponding pipe pile bearing capacity prediction model The predicted value;

[0085] Step 6.3: Use the GWO algorithm to... The update is performed to obtain the s-th candidate gray wolf individual of the (t+1)-th generation. :

[0086] Step 6.3.1: Calculate the value of the s-th gray wolf using equation (2). Each with an α wolf individual Distance between β wolf individuals Distance between δ wolf individual Distance between :

[0087] (2)

[0088] In equation (2), , , Let represent the three first coefficient vectors, and we have:

[0089] (3)

[0090] In equation (3), , , Let represent three t-th generation random vectors in the range [0,1].

[0091] Step 6.3.2: Calculate and update the s-th gray wolf individual using equation (4). ;

[0092] (4)

[0093] In equation (4), , , These represent individuals based on α wolves. β wolf individuals δ wolf individual The sth individual gray wolf after the update; , , Let represent 3 second coefficient vectors, and we have:

[0094] (5)

[0095] In equation (5), , , Let represent the other 3 t-th generation random vectors in the range [0,1]. Let t represent the update vector for the t-th generation; and The elements in the set satisfy the following conditions during the iteration process: It decreases linearly from 2 to 0.

[0096] Step 6.3.3: Use equation (6) to calculate the s-th candidate gray wolf individual in the (t+1)-th generation. ;

[0097] (6)

[0098] Step 6.4: Use the DLH algorithm to... The update is performed, resulting in the s-th candidate gray wolf individual of generation t+1. :

[0099] Step 6.4.1: Calculate using equation (7) and radius between :

[0100] (7)

[0101] In equation (7), It represents Euclidean distance.

[0102] Step 6.4.2: Calculate the s-th candidate gray wolf individual in the (t+1)-th generation using equation (8). ;

[0103] (8)

[0104] In equation (8); , respectively Centered on the radius Two randomly generated gray wolves; Let represent the random number in the t-th generation.

[0105] Step 6.5: Calculate using equation (1) and The fitness value is used to select the gray wolf individual with the smaller fitness value as the s-th gray wolf individual in the (t+1)-th generation. Thus, the (t+1)th generation wolf pack is obtained.

[0106] Step 6.6: After assigning t+1 to t, return to step 6.2 and execute sequentially until... Until then, thus obtaining the first N gray wolves in the wolf pack are selected, and the gray wolf with the lowest fitness is selected as the optimal weight and optimal threshold of the pipe pile bearing capacity prediction model after training.

[0107] Step 7: Use the trained pipe pile bearing capacity measurement model corresponding to the optimal weights and optimal thresholds as the optimal pipe pile bearing capacity prediction model to predict the pipe pile bearing capacity. The test set has 20 samples. The comparison chart of the 20 sets of predicted and actual pipe pile bearing capacity values ​​in the final test set is shown below. Figure 6 As shown, the prediction results are quite accurate.

[0108] In this embodiment, an electronic device includes a memory and a processor. The memory stores a program that supports the processor in executing the above-described method, and the processor is configured to execute the program stored in the memory.

[0109] In this embodiment, a computer-readable storage medium stores a computer program, which is executed by a processor to perform the steps of the above method.

Claims

1. A pipe pile bearing capacity prediction method based on an IGWO-BP neural network, characterized in that, The method comprises the following steps: Step 1: a numerical model of the pipe pile and soil is respectively established by using a finite element software, and a pipe pile finite element model and a soil finite element model are obtained, thereby forming a pile-soil finite element model; Step 2: the pile-soil behavior in a real situation is simulated by changing the size of the pile and the physical parameters of the soil in the pile-soil finite element model; Step 3: Obtain the static and dynamic load test data, including: the bearing capacity feature vector of the pipe pile and the displacement feature vector of the pile top wherein, respectively represent the bearing capacity feature data at the i-th static load test, represents the displacement feature data at the i-th dynamic load test; Step 3.1: The static load test is carried out on the pile-soil finite element model by applying a graded load on the pile top for calculating the load-settlement curve of the pile under the graded load, and setting the single-pile vertical compressive ultimate bearing capacity as the load value corresponding to the total settlement amount Δ according to the vertical load-settlement curve, and setting the single-pile vertical compressive bearing capacity characteristic value as half of the single-pile vertical compressive ultimate bearing capacity, so as to obtain the bearing capacity characteristic vector of the pipe pile wherein, respectively represent the bearing capacity characteristic data at the i-th static load test; T represents transposition; n is the total number of tests; Step 3.2: Create a hammer finite element model for applying an impact load to the pile, so as to carry out a dynamic load test on the pile-soil finite element model to obtain the displacement eigenvector of the pile top wherein, represents the displacement characteristic data at the i-th dynamic load test, and , represents the j-th displacement eigenvalue in , and M is the number of displacement values; Step 4: calculation the Pearson correlation coefficient between each displacement characteristic value and , and all Pearson correlation coefficients are sorted in descending order, and the displacement characteristic values corresponding to the first Pearson correlation coefficients are selected to obtain the screened displacement characteristic data of the i-th dynamic load test , thereby obtaining the screened displacement characteristic vector , wherein represents the j-th displacement characteristic value in . ​ Step 5: create a BP neural network, and take as the input of the BP neural network model, take as the output of the BP neural network model, thereby training the BP neural network, and obtaining a trained pipe pile bearing capacity prediction model; Step 6: the improved grey wolf optimization algorithm is used to optimize the weight and threshold of the trained pipe pile bearing capacity prediction model, thereby obtaining the optimal weight and optimal threshold; Step 6.1: Initialize the maximum number of iterations Define the current iteration number as t and initialize t = 1. The weight and threshold of the trained pipe pile bearing capacity prediction model are taken as the s-th gray wolf individual of the t-th generation , thereby constructing a wolf pack of the t-th generation with a size of N Step 6.2: Calculate the value of the s-th gray wolf in the t-th generation wolf pack using equation (1). fitness value The algorithm obtains and sorts all gray wolf individuals in the t-th generation wolf pack, and then selects the three gray wolf individuals with the lowest fitness values ​​as the α wolf individuals in the t-th generation wolf pack. β wolf individuals δ wolf individual ; (1) In formula (1), The s-th gray wolf individual The corresponding pipe pile bearing capacity prediction model The predicted value Step 6.3: Update the s-th candidate grey wolf individual of the t+1-th generation by using the GWO algorithm Step 6.3: Update the s-th candidate grey wolf individual of the t+1-th generation by using the GWO algorithm Step 6.3: Update the s-th candidate grey wolf individual of the t+1-th generation by using the Step 6.4: Update the s-th candidate grey wolf individual of the t+1-th generation by using the DLH algorithm ;​ Step 6.5: Calculate using equation (1) and The fitness value is used to select the gray wolf individual with the smaller fitness value as the s-th gray wolf individual in the (t+1)-th generation. Thus, the (t+1)th generation wolf pack is obtained; Step 6.6: After assigning t+1 to t, return to step 6.2 to sequentially execute until , thereby obtaining N gray wolf individuals in the t-th generation of the wolf group, and selecting the gray wolf individual with the minimum fitness from the N gray wolf individuals as the optimal weight value and the optimal threshold value of the trained pipe pile bearing capacity prediction model, respectively. ​ Step 7: the trained pipe pile bearing capacity prediction model corresponding to the optimal weight and the optimal threshold is taken as an optimal pipe pile bearing capacity prediction model, and is used for realizing the prediction of the pipe pile bearing capacity.

2. The method according to claim 1, wherein, Step 2 comprises the following steps: Step 2.1: the outer diameter, the inner diameter and the length of the pipe pile in the pipe pile finite element model are randomly changed within a certain range, so as to simulate the sizes of different pipe piles in actual engineering; Step 2.2: the elastic modulus, the cohesion and the internal friction angle of each layer of soil in the soil finite element model are randomly changed within a certain range, so as to simulate the soil quality under different geological conditions; Step 2.3: in the pile-soil finite element model, the friction coefficient between the pile-soil contact surfaces is randomly changed within a certain range, thereby obtaining the pile-soil contact behavior data in actual engineering.

3. The method according to claim 2, wherein the method is characterized by, Step 6.3 comprises the following steps: Step 6.3.1 : Calculate the distance between the s-th grey wolf individual with the formula (2) the distance between the a-th wolf individual the distance between the b-th wolf individual the distance between the a-th wolf individual the distance between the b-th wolf individual the distance between the a-th wolf individual the distance between the b-th wolf individual : (2) In formula (2), , , denote three first coefficient vectors and have: (3) In formula (3), , , denotes 3 random vectors of the t-th generation in the range [0,1]. Step 6.3.2: Calculate update for s-th grey wolf individual using formula (4) ; (4) In formula (4), , , respectively represent the updated s-th gray wolf individual based on the alpha wolf individual , the beta wolf individual , the delta wolf individual ; , , represent 3 second coefficient vectors, and have: (5) In formula (5), , , denotes a further 3rdgeneration random vector in the range [0,1]; denotes the tthgeneration update vector; and ; Step 6.3.3: Utilizing the formula (6) to count the (t+1)th generation s-th candidate gray wolf individual ; (6)。 4. The method according to claim 3, wherein, Step 6.4 comprises the following steps: Step 6.4.1 : Calculating the radius between the centers of gravity of the two circles using formula (7) and the radius : (7) In formula (7), denotes the Euclidean distance; Step 6.4.2: Calculate the s-th alternative wolf individual of the t+1 -th generation using formula (8) ; (8) In formula (8); , two gray wolf individuals are randomly generated with the center of and the radius of ; denotes the t-th random number.

5. An electronic device comprising a memory and a processor, characterized in that The memory is used for storing a program supporting the processor to execute the pipe pile bearing capacity prediction method in any one of claims 1-4, and the processor is configured to execute the program stored in the memory.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is run by the processor to execute the steps of the pipe pile bearing capacity prediction method in any one of claims 1-4.

Images