A prediction method and device for the bearing capacity of screw piles in marine soft soil areas
Through finite element numerical simulation and dimensionality reduction processing, the spiral pile bearing capacity prediction model is trained, which solves the problem of high cost and low accuracy of spiral pile bearing capacity prediction in marine soft soil areas, and achieves efficient and accurate prediction results.
Patent Information
- Application Number
- CN202410978328.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-22
- Publication Date
- 2025-07-22
- Estimated Expiration
- 2044-07-22
AI Technical Summary
The prior art has problems of high cost and low accuracy in the prediction of spiral pile bearing capacity in marine soft soil areas, and it is difficult to effectively quantify the soil and pile parameters that affect the bearing capacity.
The soil and pile variables are modeled using finite element numerical simulation method, and key variables are obtained through simulation and simulation, and the spiral pile bearing capacity prediction model is trained after dimensionality reduction, and the trained model is used for prediction.
The cost of predicting the bearing capacity of spiral piles is reduced, while improving the accuracy of prediction, reducing the dependence on soil and pile parameters, and improving the accuracy of prediction.
Smart Images
Figure CN118940431B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of foundation engineering, and particularly relates to a method and device for predicting the bearing capacity of screw piles in marine soft soil areas. Background Art
[0002] Marine soft soil is widely distributed in the southeast coastal areas of China. It has problems such as high water content and poor physical properties. The foundation structures built on it often face problems such as low bearing capacity and uneven settlement, posing huge challenges to the foundation engineering construction in coastal areas. As a special-shaped pile, a screw pile is composed of one or more spiral plates welded at equal or unequal intervals on a circular or square long screw rod or steel rod, and a certain torque is applied to the top of the screw rod by mechanical means to screw it into the soil layer. It is an anchoring structure that utilizes the resistance of deep soil. Due to its advantages such as portable installation, low pollution, and high economic efficiency, it is widely used in projects such as power grids, photovoltaic power stations, and ports in marine soft soil areas. However, screw piles also have some problems. In soft clay, the disturbance caused by installation to the surrounding soil results in insufficient bearing capacity. Increasing the length of the screw pile will lead to a large slenderness ratio, thus reducing the stability of the pile body.
[0003] Traditional prediction of the bearing capacity of screw piles often relies on empirical formulas and on-site tests. However, in actual engineering, parameters that are difficult to quantify, such as the type, density, shear strength, pile body size, and shape of the soil, will all affect the performance of the pile's bearing capacity. In the past few decades, scholars have proposed many methods for evaluating and predicting the bearing capacity of screw piles. However, the most direct and reliable method is still static loading until the pile body fails. Although there are innovative methods such as high-strain dynamic testing based on the one-dimensional wave propagation theory and with the help of advanced pile driving analyzers, due to factors such as high test costs, they have not been widely applied. Summary of the Invention
[0004] The embodiments of this application provide a method and system for predicting the bearing capacity of screw piles in marine soft soil areas, which are used to improve the prediction accuracy of the bearing capacity of screw piles while reducing the prediction cost of the bearing capacity of screw piles.
[0005] The embodiments of this invention provide a method for predicting the bearing capacity of screw piles in marine soft soil areas, and the method includes:
[0006] Obtain the soil variables in the marine soft soil area and the pile variables of the screw pile;
[0007] Use the finite element numerical simulation method to model the soil variables and the pile variables, and simulate and obtain the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively;
[0008] Based on the different values of the soil variables and the pile variables obtained from the simulation and the corresponding helical pile bearing capacities under different values, the soil variables and the pile variables are respectively dimensionally reduced to obtain key soil variables and key pile variables;
[0009] Using the different values of the key soil variables and the key pile variables and the corresponding helical pile bearing capacities under different values as sample data for model training to obtain a helical pile bearing capacity prediction model;
[0010] Predict the helical pile bearing capacity through the trained helical pile bearing capacity prediction model.
[0011] In an alternative embodiment provided by the present invention, the step of respectively dimensionally reducing the soil variables and the pile variables to obtain key soil variables and key pile variables based on the different values of the soil variables and the pile variables obtained from the simulation and the corresponding helical pile bearing capacities under different values includes:
[0012] Inputting the values of all variables in the key soil variables and the key pile variables obtained from the simulation into a feature screening model to obtain the contribution values of each variable in the key soil variables and the key pile variables for achieving the corresponding helical pile bearing;
[0013] Determine the key soil variables and the key pile variables based on the contribution values of each variable in the soil variables and the pile variables.
[0014] In an alternative embodiment provided by the present invention, the key soil variables at least include: cement soil strength and soil strength; the key pile variables at least include: helical blade diameter, pile forming diameter, number of helical blades, and helical blade spacing.
[0015] In an alternative embodiment provided by the present invention, before the simulation obtains the helical pile bearing capacities corresponding to different values of the soil variables and the pile variables, the method further includes:
[0016] Determine a first load-displacement curve through the relationship between displacement and load in the simulation;
[0017] Compare the second load-displacement curve of the in-situ compressive test with the first load-displacement curve;
[0018] If the numerical deviation between the curve comparison results is less than a preset value, the simulation obtains the helical pile bearing capacities corresponding to different values of the soil variables and the pile variables.
[0019] In an alternative embodiment provided by the present invention, before the simulated simulation obtains the corresponding screw pile bearing capacities of the soil variables and the pile variables under different value takings, the method further includes:
[0020] If the numerical deviation between the curve comparison results is less than a preset value, determine the bearing capacities corresponding to the second load-displacement curve and the first load-displacement curve respectively by the double tangent method;
[0021] If the difference in bearing capacities between the second load-displacement curve and the first load-displacement curve is less than a target value, simulate and obtain the corresponding screw pile bearing capacities of the soil variables and the pile variables under different value takings.
[0022] In an alternative embodiment provided by the present invention, the screw pile bearing capacity prediction model includes: an input layer, a hidden layer, and an output layer; the number of nodes in the input layer is the number of input variables after dimension reduction.
[0023] In an alternative embodiment provided by the present invention, using the different value takings of the key soil variables and the key pile variables and their corresponding screw pile bearing capacities under different value takings as sample data for model training to obtain a screw pile bearing capacity prediction model includes:
[0024] Perform permutations and combinations based on the number of hidden layers and the number of neurons included in the hidden layer to obtain a plurality of initial network models; and train each initial network model according to the sample data;
[0025] Calculate the mean square error of each trained initial network model;
[0026] Use the initial network model with the smallest mean square error as the screw pile bearing capacity prediction model.
[0027] In an alternative embodiment provided by the present invention, before using the different value takings of the key soil variables and the key pile variables and their corresponding screw pile bearing capacities under different value takings as sample data for model training to obtain a screw pile bearing capacity prediction model, the method further includes:
[0028] Filter out outliers and normalize the key soil variables and the key pile variables.
[0029] An embodiment of the present invention provides a screw pile bearing capacity prediction device for marine soft soil areas, and the device includes:
[0030] An acquisition module, configured to acquire soil variables in marine soft soil areas and pile variables of screw piles;
[0031] A simulation module, which is used to model the soil variables and the pile variables by using the finite element numerical simulation method, and simulate and obtain the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively;
[0032] A dimensionality reduction module, which is used to perform dimensionality reduction on the soil variables and the pile variables respectively to obtain key soil variables and key pile variables according to different values of the soil variables and the pile variables obtained by simulation and the bearing capacity of the screw pile corresponding to different values respectively;
[0033] A training module, which is used to use different values of the key soil variables and the key pile variables and the bearing capacity of the screw pile corresponding to different values respectively as sample data for model training to obtain a screw pile bearing capacity prediction model;
[0034] A prediction module, which is used to predict the bearing capacity of the screw pile through the trained screw pile bearing capacity prediction model.
[0035] The present invention provides a method and device for predicting the bearing capacity of a screw pile in a marine soft soil area. First, soil variables in the marine soft soil area and pile variables of the screw pile are obtained, and then the finite element numerical simulation method is used to model the soil variables and the pile variables, and simulate and obtain the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively; according to different values of the key soil variables and the key pile variables obtained by simulation and the bearing capacity of the screw pile corresponding to different values respectively, perform dimensionality reduction on the soil variables and the pile variables respectively to obtain key soil variables and key pile variables; use different values of the key soil variables and the key pile variables and the bearing capacity of the screw pile corresponding to different values respectively as sample data for model training to obtain a screw pile bearing capacity prediction model; finally, predict the bearing capacity of the screw pile through the trained screw pile bearing capacity prediction model. Compared with the traditional prediction of the bearing capacity of a screw pile, which often relies on empirical formulas and on-site tests, this application obtains sample data required for training the screw pile bearing capacity prediction model through simulation, and then trains the screw pile bearing capacity prediction model according to the sample data, so as to predict the bearing capacity of the screw pile according to the trained screw pile bearing capacity prediction model, thereby reducing the prediction cost of the bearing capacity of the screw pile and improving the prediction accuracy of the bearing capacity of the screw pile through this application. Description of the Drawings
[0036] Figure 1 It is a flowchart of a method for predicting the bearing capacity of a screw pile in a marine soft soil area provided by this application;
[0037] Figure 2 It is a structural diagram of a screw pile core stiffened composite pile provided by this application;
[0038] Figure 3 The network structure diagram of a prediction model for the bearing capacity of screw piles provided by this application;
[0039] Figure 4 The structural schematic diagram of a prediction device for the bearing capacity of screw piles in marine soft soil areas provided by this application. Specific embodiments
[0040] In order to better understand the above technical solutions, the technical solutions of the embodiments of this application will be described in detail below through the accompanying drawings and specific embodiments. It should be understood that the specific features in the embodiments of this application and the embodiments are detailed descriptions of the technical solutions of the embodiments of this application, rather than limitations on the technical solutions of this application. Without conflict, the technical features in the embodiments of this application and the embodiments can be combined with each other.
[0041] As Figure 1 shown, this embodiment provides a method for predicting the bearing capacity of screw piles in marine soft soil areas. The execution steps of this method are as follows:
[0042] S101, Obtain the soil variables in the marine soft soil area and the pile variables of the screw pile.
[0043] As Figure 2 shown, the screw pile in this embodiment can be a Helix Stiffened Cement Mixing Pile (HSCMP), also known as a grouting screw pile, a pressure grouting screw pile, a cement-soil screw pile, etc. By enhancing the interaction between the screw pile core and the cement soil through the screw blades, the bearing capacity of the pile foundation can be fully exerted.
[0044] In this embodiment, according to the full-scale field test method, a parameter sample of the compressive bearing capacity variables of HSCMP in the marine soft soil area is established, that is, the soil variables in the marine soft soil area and the pile variables of the screw pile are obtained. The soil variables are the soil-related data corresponding to the marine soft soil area. For example, the soil variables can include: unit weight, water content, compression modulus, ultimate side resistance, ultimate tip resistance, cement-soil strength, and soil strength, etc.; the pile variables are the relevant attribute data of the screw pile used. For example, the pile variables can include: pile stiffness, pile length, screw blade diameter, pile forming diameter, number of screw blades, and screw blade spacing, etc. This embodiment does not specifically limit the pile variables and soil variables. After the extraction of the on-site soil variables is completed, it can be as shown in Table 1 below:
[0045] Table 1
[0046]
[0047] S102. Use the finite element numerical simulation method to model the soil variables and the pile variables, and simulate and obtain the corresponding helical pile bearing capacities of the soil variables and the pile variables under different value settings.
[0048] Specifically, in this embodiment, the finite element numerical simulation method is used to model the soil variables and the pile variables. Based on the soil around the pile of the Mohr-Coulomb constitutive model and the soil hardening constitutive model, the contact surface is divided into the steel pipe-cement soil contact surface and the pile-soil contact surface according to different contact materials. The cement soil is constructed using a linear elastic non-porous material, and the pile body is based on the elastoplastic theory.
[0049] Among them, the different value settings of the soil variables and the pile variables are adjusted manually according to needs or actual situations, that is, the helical pile bearing capacities corresponding to different value settings of the soil variables and the pile variables are set under simulation.
[0050] S103. According to the different value settings of the soil variables and the pile variables obtained from the simulation and their corresponding helical pile bearing capacities under different value settings, perform dimensionality reduction on the soil variables and the pile variables respectively to obtain the key soil variables and the key pile variables.
[0051] Among them, the key soil variables are the soil variables that have a relatively large influence on the helical pile bearing capacity, and the key pile variables are the pile variables that have a relatively large influence on the helical pile bearing capacity.
[0052] In an optional embodiment provided by the present application, the step of performing dimensionality reduction on the soil variables and the pile variables respectively according to the different value settings of the soil variables and the pile variables obtained from the simulation and their corresponding helical pile bearing capacities under different value settings to obtain the key soil variables and the key pile variables includes: inputting the value settings of all variables in the key soil variables and the key pile variables obtained from the simulation into the feature screening model to obtain the contribution values of each variable in the key soil variables and the key pile variables for achieving the corresponding helical pile bearing capacity; determining the key soil variables and the key pile variables based on the contribution values of each variable in the soil variables and the pile variables. Among them, the feature screening model is a model for screening key variables, and this model can be a general neural network model. In this embodiment, a small amount of sample data can be input into this model, and then the key soil variables in the soil variables and the key pile variables in the pile variables can be determined through this model.
[0053] Specifically, in this embodiment, it can be through the formula y i = y base + f(x i1 ) + f(x i2 ) + f(x i3 ) +... + f(xij ) represents the contribution value of each sample feature in the sample data of the feature screening model to the sample label. Among them, x ij is the j-th feature in the i-th sample data; y i is the sample label (target value) of the i-th sample data; y base is the baseline of the entire data set, usually the mean of all sample target parameters, and in the application, it is the mean of the compressive ultimate bearing capacity of all samples. f(x ij ) is the contribution value of the j-th feature in the i-th sample data to the target value y i .
[0054] In this embodiment, after dimension reduction of the soil variables and pile variables by the above method, the obtained key soil variables at least include: cement soil strength and soil strength; the key pile variables at least include: auger blade diameter, pile forming diameter, number of auger blades, and auger blade spacing.
[0055] Furthermore, in order to ensure the accuracy of the bearing capacity of the screw pile obtained by simulation in this embodiment, it is also necessary to verify the result of the simulation. That is, before obtaining the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables through simulation, it is necessary to determine the first load-displacement curve through the relationship between displacement and load in the simulation; compare the second load-displacement curve of the on-site compressive test with the first load-displacement curve; if the numerical deviation between the curve comparison results is less than the preset value, the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables is obtained through simulation. The abscissa in the load-displacement curve is the bearing capacity, and the ordinate is the displacement.
[0056] To further verify the accuracy of the simulation result, before obtaining the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables through simulation in this embodiment, the method further includes: if the numerical deviation between the curve comparison results is less than the preset value, determine the bearing capacities corresponding to the second load-displacement curve and the first load-displacement curve by the double-tangent method; if the difference between the bearing capacities of the second load-displacement curve and the first load-displacement curve is less than the target value, the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables is obtained through simulation. Among them, the preset value and the target value are values set according to actual needs, and this embodiment does not make specific limitations on this.
[0057] It should be noted that the double-tangent method is to take the intersection points of two tangents of the gentle part and the end part of the curve on the first load-displacement curve and the second load-displacement curve respectively, and then compare the bearing capacities of the two intersection points. If the difference in the bearing capacities of the two intersection points is less than the target value, the simulated soil variables and the pile variables corresponding to different values are obtained through simulation. For example, the ultimate bearing capacity of the curve is determined by the double-tangent method. The ultimate bearing capacity and displacement of the first load-displacement curve are 504.19 KN and 22.7 mm, and those of the second load-displacement curve are 494.17 KN and 21.4 mm, and the deviation of the ultimate bearing capacity is 2.03%. The above results show that the simulation model can better simulate and restore the actual deformation of the HSCM pile in soft soil under the action of downward pressure.
[0058] S104, using the different values of the key soil variables and the key pile variables and the corresponding screw pile bearing capacities under different values as sample data for model training to obtain a screw pile bearing capacity prediction model.
[0059] In this embodiment, before using the different values of the key soil variables and the key pile variables and the corresponding screw pile bearing capacities under different values as sample data for model training to obtain a screw pile bearing capacity prediction model, the method further includes: filtering outliers and normalizing the key soil variables and the key pile variables. In this embodiment, normalization can be performed through the following formula:
[0060]
[0061] where X represents the original data, X min and X max represent the minimum value and the maximum value in the sample data set respectively. Through this conversion method, it is ensured that all eigenvalues will be in the range of 0 to 1, thus eliminating the influence of the dimension and unifying the data scale.
[0062] It should be noted that the helical pile bearing capacity prediction model in this embodiment includes an input layer, a hidden layer, and an output layer; the number of nodes in the input layer is the number of input variables after dimension reduction. The training process of this helical pile bearing capacity prediction model is as follows: multiple initial network models are obtained by arranging and combining based on the number of hidden layers and the number of neurons in the hidden layer; and each initial network model is trained according to the sample data; calculate the mean square error of each trained initial network model; the initial network model with the smallest average error is used as the helical pile bearing capacity prediction model. For example, if the hidden layer contains at most 3 layers, and the number of neurons in each hidden layer is 5 - 10, then permutations and combinations can be performed based on these data, and then each initial network model is trained according to the sample data; calculate the mean square error of each trained initial network model; the initial network model with the smallest average error is used as the helical pile bearing capacity prediction model. For example, by comparison, it is found that the initial network model with 3 hidden layers and 10 neurons in each hidden layer has the smallest mean square error, then the network structure of the initial network model is used as the network structure of the helical pile bearing capacity prediction model.
[0063] For example, in this embodiment, a small amount of sample data can be used to train multiple initial network models, and then one is selected from the multiple trained initial network models as the helical pile bearing capacity prediction model. Each group consists of input parameters (sample data, that is, the values corresponding to the key soil variables and key pile variables) and output parameters (sample labels corresponding to the sample data, that is, the helical pile bearing capacity obtained through simulation). The sample data can specifically be blade spacing, number of blades, cement soil strength, soil strength, pile forming diameter, helical blade diameter, etc.
[0064] After the network structure of the helical pile bearing capacity prediction model in this embodiment, the helical pile bearing capacity prediction model is continuously trained with other sample data until the number of iterations reaches the maximum value or the error is within the preset error range, and the model training is completed. Among them, the error can be determined by means such as mean square error MSE and correlation.
[0065] As Figure 3 shown is the network structure diagram of a helical pile bearing capacity prediction model provided in this embodiment. The calculation process of each neuron in the hidden layer of this helical pile bearing capacity prediction model is as follows:
[0066]
[0067] y j = Φ(h j ) = Φ(w yj h j + b j2 );
[0068] Among them, n is the number of input variables, that is, the total number of key soil variables and key pile variables, w ij is the weight of the i-th variable in the j-th neuron, x i is the i-th input variable. For example, x1 represents the number of spiral blades, and x2 represents the pitch of the spiral blades; b j1 is the threshold between the input layer and the first hidden layer, f is the activation function, Φ is the linear function, w yj is the weight between the j-th neuron and the output variable y, b j2 is the threshold between the first hidden layer and the second hidden layer. In this embodiment, the bearing capacity prediction model of the screw pile regards the input variables as nodes one by one. For example, this application has 6 input variables, that is, 6 nodes. The code names of these nodes are i, x i value is the specific value (for example, for input node 1, the corresponding x1 value is 5). Similarly, j represents each neuron in the next layer, that is, the first hidden layer.
[0069] For example, in this embodiment, the neural network structure of the screw pile bearing capacity prediction model is: input layer - first hidden layer - second hidden layer - third hidden layer - fourth hidden layer - output layer. The input layer includes six nodes, and each of the four hidden layers contains 10 neurons. The output layer has one node, that is, the bearing capacity of the screw pile. In this embodiment, there is a threshold between each hidden layer, b j1 refers to the threshold between the input layer and the first hidden layer, b j2 refers to the threshold between the first hidden layer and the second hidden layer. This threshold is a concept in the neural network and can be simply understood as a deviation. A value multiplied by the weight and then added with the deviation is the value finally output by this neuron. Then this value is told to the next hidden layer neuron, and the next neuron also performs the same calculation, and so on, until the final output parameter is obtained.
[0070] Taking the input layer and the first hidden layer as an example, the nodes in the input layer tell the values of the input variables to each neuron in the hidden layer. The neuron with the code name j needs to perform three operations: summation, activation, and mapping. Summation is the first formula Activation is the second formula Mapping is the third formula y j =Φ(h j )=Φ(w yj h j +b j2 ). Then this neuron tells the y j obtained by mapping to the next hidden layer, and the neurons in the next hidden layer perform the same operation, and so on, until the output parameter, that is, the compressive bearing capacity, is finally obtained.
[0071] S105. Predict the bearing capacity of the screw pile through the trained bearing capacity prediction model of the screw pile.
[0072] Specifically, after the bearing capacity prediction model of the screw pile is trained in this embodiment, the pile body variables of the screw pile to be detected and the soil body variables in the marine soft soil area where it is located can be extracted to obtain the key pile body variables and key soil body variables, and then the key pile body variables and key soil body variables are input into the bearing capacity prediction model of the screw pile to obtain the bearing capacity of the screw pile to be detected.
[0073] A method for predicting the bearing capacity of a screw pile in a marine soft soil area provided by an embodiment of the present application first obtains the soil body variables in the marine soft soil area and the pile body variables of the screw pile, and then uses the finite element numerical simulation method to model the soil body variables and pile body variables, and simulates and obtains the bearing capacity of the screw pile corresponding to different values of the soil body variables and pile body variables respectively; according to the different values of the key soil body variables and key pile body variables obtained by the simulation and the bearing capacity of the screw pile corresponding to different values respectively, the soil body variables and pile body variables are respectively reduced in dimension to obtain the key soil body variables and key pile body variables; the different values of the key soil body variables and key pile body variables and the bearing capacity of the screw pile corresponding to different values respectively are used as sample data for model training to obtain the bearing capacity prediction model of the screw pile; finally, the bearing capacity of the screw pile is predicted through the trained bearing capacity prediction model of the screw pile. Compared with the traditional prediction of the bearing capacity of the screw pile, which often relies on empirical formulas and field tests, the present application obtains the sample data required for training the bearing capacity prediction model of the screw pile through simulation, and then trains the bearing capacity prediction model of the screw pile according to the sample data, so as to predict the bearing capacity of the screw pile according to the trained bearing capacity prediction model of the screw pile, thereby reducing the prediction cost of the bearing capacity of the screw pile and improving the prediction accuracy of the bearing capacity of the screw pile through the present application.
[0074] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.
[0075] In one embodiment, a device for predicting the bearing capacity of a screw pile in a marine soft soil area is provided. The device for predicting the bearing capacity of a screw pile in a marine soft soil area corresponds one-to-one with the method for predicting the bearing capacity of a screw pile in a marine soft soil area in the above embodiment. As Figure 4 shown, the detailed description of each functional module of the device for predicting the bearing capacity of a screw pile in a marine soft soil area is as follows:
[0076] An acquisition module 41, configured to acquire the soil body variables in the marine soft soil area and the pile body variables of the screw pile;
[0077] A simulation module 42, which is used to model the soil variables and the pile variables by using the finite element numerical simulation method, and simulate and obtain the bearing capacity of the screw pile corresponding to the soil variables and the pile variables respectively under different value-taking conditions;
[0078] A dimensionality reduction module 43, which is used to respectively perform dimensionality reduction on the soil variables and the pile variables according to the different value-taking conditions of the soil variables and the pile variables obtained from the simulation and the bearing capacity of the screw pile corresponding to them respectively under different value-taking conditions, so as to obtain key soil variables and key pile variables;
[0079] A training module 44, which is used to use the different value-taking conditions of the key soil variables and the key pile variables and the bearing capacity of the screw pile corresponding to them respectively under different value-taking conditions as sample data for model training to obtain a screw pile bearing capacity prediction model;
[0080] A prediction module 45, which is used to predict the bearing capacity of the screw pile through the trained screw pile bearing capacity prediction model.
[0081] In an optional embodiment provided by the present invention, the dimensionality reduction module 43 is specifically used for:
[0082] Input the value-taking of all variables in the key soil variables and the key pile variables obtained from the simulation into a feature screening model, and obtain the contribution value of each variable in the key soil variables and the key pile variables when obtaining the corresponding screw pile bearing;
[0083] Determine the key soil variables and the key pile variables based on the contribution values of each variable in the soil variables and the pile variables.
[0084] In an optional embodiment provided by the present invention, the key soil variables at least include: the strength of the cement soil and the strength of the soil; the key pile variables at least include: the diameter of the screw blade, the pile-forming diameter, the number of screw blades, and the pitch of the screw blades.
[0085] In an optional embodiment provided by the present invention, the simulation module 42 is specifically used for:
[0086] Determine a first load-displacement curve through the relationship between displacement and load in the simulation;
[0087] Compare the second load-displacement curve of the on-site compressive test with the first load-displacement curve;
[0088] If the numerical deviation between the curve comparison results is less than a preset value, simulate and obtain the bearing capacity of the screw pile corresponding to the soil variables and the pile variables respectively under different value-taking conditions.
[0089] In an alternative embodiment provided by the present invention, the simulation module 42 is specifically configured to:
[0090] If the numerical deviation between the curve comparison results is less than a preset value, determine the bearing capacities corresponding to the second load-displacement curve and the first load-displacement curve respectively by the double tangent method;
[0091] If the bearing capacity difference between the second load-displacement curve and the first load-displacement curve is less than a target value, simulate and obtain the corresponding screw pile bearing capacities when the soil variables and the pile variables take different values.
[0092] In an alternative embodiment provided by the present invention, the screw pile bearing capacity prediction model includes: an input layer, a hidden layer, and an output layer; the number of nodes in the input layer is the number of input variables after dimension reduction.
[0093] In an alternative embodiment provided by the present invention, the training module 44 is specifically configured to:
[0094] Perform permutations and combinations based on the number of hidden layers and the number of neurons included in the hidden layer to obtain a plurality of initial network models; and train each initial network model according to the sample data respectively;
[0095] Calculate the mean square error of each trained initial network model;
[0096] Take the initial network model with the smallest mean square error as the screw pile bearing capacity prediction model.
[0097] In an alternative embodiment provided by the present invention, the calculation process of each neuron in the hidden layer is:
[0098]
[0099] y = Φ(h j ) = Φ(w yj h j + b j2 );
[0100] Where n is the number of input variables, w ij is the weight of the i-th variable in the j-th neuron, x i is the i-th input variable, b j1 is the threshold between the input layer and the first hidden layer, f is the activation function, Φ is the linear function, w yj is the weight between the j-th neuron and the output variable y, and b j2 is the threshold between the first hidden layer and the second hidden layer.
[0101] In an alternative embodiment provided by the present invention, before using different values of the key soil variables and the key pile variables and their corresponding auger pile bearing capacities under different values as sample data for model training to obtain an auger pile bearing capacity prediction model, the method further includes:
[0102] Filtering outliers and normalizing the key soil variables and the key pile variables.
[0103] For the specific limitations on the auger pile bearing capacity prediction device for marine soft soil areas, reference can be made to the limitations on the auger pile bearing capacity prediction method for marine soft soil areas described above, which will not be elaborated here. Each module in the above device can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor of the computer device in hardware form or be independent of it, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to the above modules.
[0104] Those skilled in the art can clearly understand that for the convenience and brevity of description, only the above division of each functional unit and module is used as an example. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above.
[0105] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A method for predicting the bearing capacity of screw piles in marine soft soil areas, characterized in that, The method includes: S101, obtaining the soil variables in the marine soft soil area and the pile variables of the screw pile; S102, using the finite element numerical simulation method to model the soil variables and the pile variables, and simulating to obtain the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively; S103, according to the different values of the soil variables and the pile variables obtained by simulation and the bearing capacity of the screw pile corresponding to different values respectively, reducing the dimensions of the soil variables and the pile variables to obtain the key soil variables and the key pile variables; S104, using the different values of the key soil variables and the key pile variables and the bearing capacity of the screw pile corresponding to different values respectively as sample data for model training to obtain a screw pile bearing capacity prediction model; S105, predicting the bearing capacity of the screw pile through the trained screw pile bearing capacity prediction model; Before the simulation obtains the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively, the method further includes: Determining a first load-displacement curve through the relationship between displacement and load in the simulation; comparing the second load-displacement curve of the on-site compressive test with the first load-displacement curve; if the numerical deviation between the curve comparison results is less than a preset value, simulating to obtain the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively; if the numerical deviation between the curve comparison results is less than a preset value, determining the bearing capacity corresponding to the second load-displacement curve and the first load-displacement curve respectively by the double-tangent method; if the bearing capacity difference between the second load-displacement curve and the first load-displacement curve is less than the target value, simulating to obtain the bearing capacity of the screw pile corresponding to different values of the soil variables and the pile variables respectively; S103 includes: Inputting the values of all variables in the key soil variables and the key pile variables obtained by simulation into a feature screening model to obtain the contribution value of each variable in the key soil variables and the key pile variables to obtaining the corresponding screw pile bearing; determining the key soil variables and the key pile variables based on the contribution value of each variable in the soil variables and the pile variables; wherein, the feature screening model is a model for screening key variables, and this model is a general neural network model. By inputting a small amount of sample data into this model, then determining the key soil variables in the soil variables and the key pile variables in the pile variables through this model; In S104, the screw pile bearing capacity prediction model includes: an input layer, a hidden layer, and an output layer; the number of nodes in the input layer is the number of input variables after dimension reduction; The step of using the different values of the key soil variables and the key pile variables and the bearing capacity of the screw pile corresponding to different values respectively as sample data for model training to obtain a screw pile bearing capacity prediction model includes: Arrange and combine based on the number of hidden layers and the number of neurons in the hidden layer to obtain multiple initial network models; and train each initial network model according to the sample data respectively; Calculate the mean square error of each trained initial network model; Use the initial network model with the smallest average error as the helical pile bearing capacity prediction model.
2. The method according to claim 1, wherein The key soil variables at least include: cement soil strength and soil strength; the key pile variables at least include: helical blade diameter, pile forming diameter, number of helical blades and helical blade spacing.
3. The method according to claim 1, wherein Before using the different values of the key soil variables, the key pile variables and the corresponding helical pile bearing capacities under different values as sample data for model training to obtain a helical pile bearing capacity prediction model, the method further includes: Filter out outliers and normalize the key soil variables and the key pile variables.
4. A bearing capacity prediction device for helical piles in marine soft soil areas, which implements the method described in claim 1, characterized in that, The device includes: An acquisition module for acquiring soil variables in marine soft soil areas and pile variables of helical piles; A simulation module for modeling the soil variables and the pile variables by using the finite element numerical simulation method, and simulating to obtain the helical pile bearing capacities corresponding to the different values of the soil variables and the pile variables respectively; A dimensionality reduction module for respectively performing dimensionality reduction on the soil variables and the pile variables according to the different values of the soil variables and the pile variables obtained by simulation and the corresponding helical pile bearing capacities under different values to obtain key soil variables and key pile variables; A training module for using the different values of the key soil variables, the key pile variables and the corresponding helical pile bearing capacities under different values as sample data for model training to obtain a helical pile bearing capacity prediction model; A prediction module for predicting the helical pile bearing capacity through the trained helical pile bearing capacity prediction model.
Citation Information
Patent Citations
Pile-soil interaction prediction analysis method based on machine learning
CN111209708A