A method for predicting pullout capacity of suction buckets

Abstract

The application provides a suction bucket uplift bearing capacity prediction method, comprising the following steps: establishing a suction anchor uplift visualization experiment device; performing suction anchor uplift experiments under different working conditions, and obtaining the relationship curve of the uplift tension and displacement under different working conditions through data processing; generating a suction bucket "factor-uplift force" database; selecting a BP-RF machine learning algorithm for small sample data prediction; comparing and analyzing the prediction results and the measured results, and judging the influence law of the uplift force. The application considers the limitations of the experiment device and conditions, and establishes a BP-RF hybrid machine learning method for small sample databases by comprehensively considering the advantages of the BP neural network algorithm and the RF algorithm in processing small sample problems. The suction bucket uplift bearing capacity prediction model is established based on the method, so that the influence law of the length, diameter and uplift rate of the suction bucket on the uplift bearing capacity can be comprehensively and accurately evaluated, and the design of the uplift bearing capacity of the suction bucket is referred to.

Description

Technical Field

[0001] This invention relates to the field of suction bucket foundation technology, and in particular to a method for predicting the upward pull-out load-bearing capacity of a suction bucket. Background Technology

[0002] Offshore wind farms will expand into deeper waters in the future, and floating wind turbines are an inevitable choice for future development. Offshore wind power will inevitably move from nearshore to open ocean, and from shallow to deep sea. Floating wind turbines are connected to deep-sea suction anchors via anchor chains to withstand the uplift force caused by wind and waves. The marine environment is complex and changeable, often accompanied by severe typhoons and complex sea conditions. When storm surges occur, the overturning moment of offshore wind turbine foundations is very large, and the corresponding uplift force transmitted from the anchor cables to the suction anchor is also very large. Therefore, it is essential to study and determine the ultimate uplift bearing capacity of suction anchors, providing a theoretical basis for the installation and design of wind turbine foundations in deep seas. Studying the mechanical characteristics of deep-sea suction anchors under storm surges has considerable engineering significance. Although a lot of research has been conducted on the uplift bearing characteristics of suction anchors, few studies have focused on the deformation of the soil inside and outside the suction anchor during the uplift process. Moreover, the uplift rate is very slow when conducting uplift experiments on suction anchors in sandy soil, and most studies focus on the bearing characteristics of suction anchors under static loads. There is very little research on its uplift bearing characteristics under high-speed loads, especially under completely undrained conditions. Summary of the Invention

[0003] The purpose of this invention is to overcome the defects of the prior art and provide a method for predicting the pull-out bearing capacity of a suction anchor, enabling the study of the influence of the length, diameter, and pull-out rate of the suction barrel on the pull-out bearing capacity.

[0004] To achieve the above objectives, this invention proposes a method for predicting the upward pull-out load-bearing capacity of a suction bucket, comprising the following steps:

[0005] S1: Establish a visual experimental device for pulling up a suction anchor;

[0006] S2: Using the experimental device, conduct suction anchor pull-out experiments under different working conditions, and obtain the relationship curve of pull-out force versus displacement under different working conditions through data processing.

[0007] S3: Generate a database of "factors - upward force" for the suction bucket based on the relationship curve of upward pulling force with displacement under different working conditions;

[0008] S4: Use the BP-RF machine learning algorithm for small sample data prediction;

[0009] S5: Compare and analyze the predicted results and the measured results, and determine the influence law of the upward pulling force.

[0010] Furthermore, in step S1, the suction anchor pull-up visualization experimental device includes a suction anchor system, a reaction frame, and a control mechanism:

[0011] The suction anchor system includes a hollow cylindrical suction anchor that is closed at the top and open at the bottom. A pull rod is connected to the center of the anchor cap at the top of the hollow cylindrical anchor, and the pull rod is connected to a drive mechanism. A tension sensor is installed on the pull rod, and the tension sensor is connected to a tension sensor data acquisition instrument. The anchor cap is also provided with a drainage and venting hole with a valve to discharge water and air trapped in the anchor during the sinking process. Before pulling the suction anchor up, the valve is closed to seal the drainage and venting hole.

[0012] The hollow cylindrical barrel is fixed inside a glass box filled with sand. The glass box is located within the reaction frame. The control mechanism is signal-connected to the suction anchor system.

[0013] The tension sensor, reaction frame, and drive mechanism together constitute the loading system.

[0014] Furthermore, the sand is transparent soil; the reaction frame is welded from square steel pipes, serving to stabilize and transmit reaction force; the drive mechanism is a linear servo electric cylinder, used to stably provide different speeds to drive the lever movement, with a lever lifting range of 0~300mm, a pulling force range of 0~1000N, and a moving speed range of 1~500mm / s, the moving speed and lifting position can be adjusted via the touch panel on the control box; the control mechanism is a computer.

[0015] Furthermore, a threaded hole is provided in the center of the anchor cap for threaded connection of the pull rod.

[0016] Furthermore, the linear servo electric cylinder is connected to a guide rod, which is fixed to the top of the reaction frame and located above the suction barrel, for controlling the relative position of the suction barrel in the sand.

[0017] Furthermore, step S2 specifically involves:

[0018] S2.1: Design working condition, wherein the parameters that can be changed include the inner diameter of the suction barrel, the outer diameter of the suction barrel, the length of the suction barrel, and the upward force loading rate;

[0019] S2.2: Fill the prepared sand and soil into the plexiglass box;

[0020] S2.3: Using the drive mechanism, the suction anchor is driven into the sand inside the glass box;

[0021] S2.4: After the suction anchor penetrates the sand, let it stand for a period of time;

[0022] S2.5: Pull the suction anchor upwards and record the reading from the tension sensor;

[0023] S2.6: Using the control mechanism, data processing is performed to further transform the curve of the tension change over time on the tension sensor into the relationship curve of the pull-out tension change over displacement.

[0024] Furthermore, step S2.3 specifically involves: first, opening the drainage valve on the suction anchor cover and adjusting the position of the model box at the bottom of the suction anchor so that the suction anchor is at the center of the plane of the model box; second, using the drive mechanism to insert the suction anchor into the designated position in the sand.

[0025] Furthermore, in step S2.5, the method of pulling the suction anchor is displacement control. The pulling speed is controlled by the drive mechanism to pull the suction anchor up at different loading rates, and the reading of the tension sensor is recorded by the control mechanism.

[0026] Furthermore, in step S3, the factors include: the barrel length of the cylindrical suction anchor, the barrel diameter of the cylindrical suction anchor, and the upward pull rate.

[0027] Furthermore, step S4 specifically involves:

[0028] S4.1: Establish an input-output relationship table, setting different factors as... This forms the input matrix X;

[0029] S4.2: Map the input-output relationship table to a BP neural network model, and learn a mapping through training. Predict the output target;

[0030] S4.3: Map the input-output relationship table to a random forest (RF) model, and learn by training multiple decision trees; Predict the output target;

[0031] S4.4: The prediction results of the BP neural network model and the random forest model are weighted and fused to obtain the final prediction result.

[0032] Furthermore, the mathematical model of the BP neural network is as follows:

[0033] ;

[0034] Input vectors, representing different factors in the database;

[0035] , Weight matrix;

[0036] , Bias vector;

[0037] Activation function (e.g., ReLU);

[0038] Output layer function (e.g., linear activation);

[0039] Furthermore, the mathematical model for Random Forest (RF) is as follows:

[0040] ;

[0041] Represents the i-th decision tree pair with input. The predicted value; N is the total number of decision trees.

[0042] Furthermore, in step S4.4, the weighted fusion formula is:

[0043] ;

[0044] In the formula, Mixed weights reflect the contributions of both models and are adjusted. The optimal value can be found through a validation set; The prediction result is from the BP neural network; This represents the prediction results from the random forest.

[0045] Furthermore, in step S5, the predicted results and the measured results are compared to analyze the influence trend of different factors on the upward pulling force, and the influence law of the upward pulling force is judged by combining image analysis and machine learning methods.

[0046] Furthermore, the suction anchor pull-up visualization experimental device also includes a camera and a laser mechanism located outside the reaction frame. The camera is located outside the reaction frame and faces the glass box. The camera is connected to the computer via a cable. The laser mechanism includes a laser emitter, a linear lens, and a laser power supply. The power supply is connected to the laser emitter via a cable. The laser emitter faces the linear lens, and the linear lens faces the glass box.

[0047] Compared with the prior art, the advantages of the present invention are:

[0048] 1. Considering the limitations of experimental equipment and conditions, this invention, targeting small sample databases, combines the advantages of BP neural network algorithm and RF algorithm for handling small sample problems, and establishes a BP-RF hybrid machine learning method. Based on this method, a model for predicting the pull-out bearing capacity of the suction bucket is established, thereby exploring the influence of the length, diameter and pull-out rate of the suction bucket on the pull-out bearing capacity, and providing a reference for the design of the pull-out bearing capacity of the suction bucket.

[0049] 2. This invention uses a hybrid machine learning method to comprehensively consider various factors to predict the upward bearing capacity, and uses a single variable to explain the upward bearing capacity, explaining the law of the upward bearing capacity of the suction bucket from the perspective of single factor and multi-factor coupling.

[0050] 3. This invention establishes a visual experimental device for the upward pull of suction anchors to conduct experiments on the upward pull of suction anchors under different factors. The experimental data is used as a machine learning sample library, which avoids the disadvantage of the prediction results not matching reality caused by using numerical simulation results as a sample library. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the method for predicting the upward pull-out bearing capacity of the suction bucket proposed in an embodiment of the present invention.

[0052] Figure 2 This is a schematic diagram of the suction anchor pull-out visualization test device in an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of the cross-sectional structure of the suction barrel of the suction anchor in an embodiment of the present invention;

[0054] Figure 4 This is a comparison chart of BP-RF prediction results obtained by the method of this invention and measured data;

[0055] Figure 5 This is a graph showing the relationship between the pull-out force on the suction anchor and the loading rate obtained through the method of this embodiment of the invention.

[0056] Figure 6 This is a graph showing the relationship between the pull-out force and the length-to-diameter ratio of the suction anchor obtained by the method of this embodiment of the invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be further described below.

[0058] This embodiment proposes a method for predicting the upward pull-out load-bearing capacity of a suction bucket, such as... Figure 1 As shown, the method includes the following steps:

[0059] S1: Establish a visual experimental device for pulling up a suction anchor;

[0060] The suction anchor pull-out visualization experimental device includes a suction anchor system and a loading system, such as... Figure 2 As shown, the loading system includes a tension sensor 5, a reaction frame 9, and a linear servo electric cylinder 6. The suction anchor system includes a hollow thin-walled cylindrical suction anchor 2, which is closed at the top and open at the bottom. This hollow cylindrical suction anchor 2 is made of a series of acrylic plexiglass components, such as... Figure 3As shown, the top anchor cap is equipped with a level 23 and a drain and vent hole 21 with a valve, used to discharge water and air trapped inside the anchor during the sinking process. Before pulling up the suction anchor 2, the valve is closed to seal the drain and vent hole 21. In addition, an 8mm diameter threaded hole 22 is drilled in the center of the anchor cap, through which a pull rod is connected. The pull rod is connected to a linear servo electric cylinder 6 via a flange 4. The linear servo electric cylinder 6 is used to stably provide different speeds to drive the pull rod. The pull rod lifting range is 0~300mm, the tension range is 0~1000N, and the moving speed range is 1~500mm / s. The moving speed and lifting position can be adjusted via the touch panel on the control box. The tension sensor 5 is fixedly installed between the pull rod and the output end of the linear servo electric cylinder 6. The tension sensor 5 is also connected to a tension sensor data acquisition instrument 8 via a cable to collect tension values.

[0061] A hollow cylindrical suction anchor 2 is located inside an acrylic plexiglass box 1, which is filled with transparent soil 3. A linear servo electric cylinder 6 is used to insert the hollow cylindrical suction anchor 2 into the transparent soil 3 inside the acrylic plexiglass box 1. The linear servo electric cylinder 6 is connected to a guide rod 7. Figure 2 As shown, the guide rod 7 is fixed to the top of the reaction frame 9, located above the suction anchor 2, and is used to control the relative position of the hollow cylindrical suction anchor 2 in the transparent soil 3. The glass box 1 is placed inside the metal reaction frame 9. In this embodiment, the reaction frame 9 is welded from square steel pipes and plays the role of stabilizing and transmitting the reaction force; the suction anchor system is connected to the microcomputer 11 via signal.

[0062] In addition, a camera 10 and a laser mechanism are also provided outside the reaction frame 9. The camera 10 is located on the outside side of the reaction frame 9, facing the side of the glass box 1, and is connected to the microcomputer 11 via a cable. The laser mechanism includes a laser emitter 12, a linear lens 13, and a laser power supply group 14, such as... Figure 2 As shown, the laser power supply unit 14 is connected to the laser emitter 12 by cable. The laser emitter 12 faces the line lens 13, which faces the front of the glass box 1.

[0063] In this embodiment, the internal clear dimensions of the acrylic plexiglass box 1 are 18cm × 18cm × 40cm, and the wall thickness is 1cm.

[0064] In this embodiment, the five suction anchors 2 selected are denoted as M-1, M-2, M-3, M-4, and M-5, respectively. Their dimensional parameters, such as average diameter D, length L, side wall thickness t1, and top wall thickness t2 of the suction barrel, are shown in Table 1.

[0065] Table 1. Selected suction anchor parameters

[0066]

[0067] S2: Using the above experimental setup, pull-out experiments of the suction anchor under different working conditions were conducted, and the relationship curves between the pull-out force and displacement under different working conditions were obtained through data processing. Specifically:

[0068] S2.1: Design condition, the parameters that can be changed in this condition include the inner diameter of the suction barrel, the outer diameter of the suction barrel, the length of the suction barrel, and the upward force loading rate;

[0069] S2.2: Fill the acrylic plexiglass box 1 with the prepared transparent soil 3;

[0070] S2.3: Using a linear servo electric cylinder 6, the suction anchor 2 is inserted into the transparent soil 3 inside the glass box. The specific process of this step is as follows: First, open the drain valve on the anchor cover of the suction anchor 2 and adjust the position of the glass box 1 at the bottom of the suction anchor so that the suction anchor 2 is at the center of the plane of the glass box. Second, use the linear servo electric cylinder 6 to insert the suction anchor 2 into the designated position inside the transparent soil 3 inside the glass box 1.

[0071] S2.4: After the suction anchor 2 penetrates the transparent soil 3, let it stand for a period of time;

[0072] S2.5: After standing still, pull the suction anchor 2 upward and record the reading of the tension sensor; wherein, the method of pulling the suction anchor 2 is displacement control, and the upward speed is controlled by the linear servo electric cylinder 6, so that the suction anchor 2 is pulled upward at different loading rates, and the reading of the tension sensor is collected by the tension sensor data acquisition instrument 8 and transmitted to the microcomputer 11.

[0073] In this embodiment, the tension sensor 5 has a range of 500N and an accuracy of 0.01N, and the tension data acquisition instrument 8 has a sampling frequency of 10Hz. The loading rates of the above five suction anchors are recorded in Table 2 below:

[0074] Table 2. Record of Uplift Loading Rate for Anchors with Different Suction Forces

[0075]

[0076] S2.6: Using the microcomputer 11, data processing is performed to further transform the curve of the tension change over time on the tension sensor into the curve of the relationship between the pull-out tension and the displacement.

[0077] S3: Based on the relationship curve of the pull-out force with displacement under different working conditions, generate a database of "factors - pull-out force" for the suction bucket. The specific factors include: the length of the cylindrical suction anchor, the diameter of the cylindrical suction anchor, and the pull-out rate.

[0078] In this embodiment, a portion of the database is shown in Table 3 below:

[0079] Table 3. "Factors - Uplift Force" Database (Partial)

[0080]

[0081] S4: The BP-RF (Neural Network-Random Forest) machine learning algorithm is selected for prediction of small sample data. The method is as follows:

[0082] S4.1: Establish an input-output relationship table, setting different factors as... This forms the input matrix X;

[0083] S4.2: Map the above input-output relationship table to the BP neural network model. The mathematical model of the BP neural network is as follows:

[0084] ;

[0085] Input vectors, representing different factors in the database;

[0086] , Weight matrix;

[0087] , Bias vector;

[0088] Activation function (e.g., ReLU);

[0089] Output layer function (e.g., linear activation);

[0090] Learn a mapping through training. Predict the output target;

[0091] S4.3: Map the above input-output relationship table to a Random Forest (RF) model. The mathematical model of the Random Forest (RF) is as follows:

[0092] ;

[0093] Represents the i-th decision tree pair with input. The predicted value; N is the total number of decision trees.

[0094] Learning by training multiple decision trees Predict the output target;

[0095] S4.4: The prediction results from the BP neural network model and the random forest model are weighted and fused to obtain the final prediction result. The weighted fusion formula is:

[0096] ;

[0097] In the formula, Mixed weights reflect the contributions of both models and are adjusted. The optimal value can be found through a validation set; The prediction result is from the BP neural network; This represents the prediction results from the random forest.

[0098] S5: Compare the predicted results with the measured results, analyze the influence trend of different factors on the upward force, and use image analysis and machine learning methods to comprehensively judge the influence law of the upward force.

[0099] In this embodiment, the comparison between the predicted results and the measured results is as follows: Figure 4 As shown, from Figure 4 As can be seen, the closer the scatter points are to the ideal line, the more accurate the prediction results of the BP-RF model.

[0100] In this embodiment, the evaluation index of the calculation model is calculated. The mean of the sum of squares of the errors between the predicted and actual values ​​is calculated as MSE = 5579, where R0 2 The proportion of the variance of the target variable explained by the model is measured from The calculation results show that the model has high accuracy. The excessively large MSE value is due to the large fluctuations in the measured data, the small data sample size, and the fact that the predicted value y itself is on the order of magnitude larger than the actual value. At the level of [level missing], therefore, from an experimental perspective, considering the large number of model parameters, the BP-RF model can predict the upward pull-out bearing capacity of the suction bucket caused by different factors relatively well. The following section will explain the influence trend of different factors on the upward pull-out bearing capacity from the perspective of controlling a single variable, using image analysis.

[0101] The influence of the anchor length on the pull-out force of a suction anchor is as follows: Figure 5 As shown in (a), the influence of the suction anchor diameter (the cross-sectional area of ​​the suction anchor) on the pull-out force is as follows: Figure 5 As shown in (b), the patterns are significantly different. Relatively speaking, the anchor diameter (cross-sectional area of ​​the anchor) has a more significant impact on the bearing capacity of the suction anchor foundation. When the pull-out rate is low (v=1mm / s), this difference in the influence of anchor length and anchor diameter on bearing capacity is not obvious. However, when the anchor is pulled out at high speed, the anchor diameter plays a dominant role in bearing capacity, that is, the negative pressure at the top provides most of the pull-out resistance.

[0102] Figure 6 (a) and Figure 6 Figure (b) shows the curves of the uplift force versus the loading rate for different suction anchor length-diameter ratios L / D:

[0103] Figure 6 Figure (a) shows the relationship between the uplift force and the uplift rate when the diameter of the suction anchor is fixed at D = 50 and the length-diameter ratios L / D are 0.5, 1, and 1.5 respectively. Figure 6 As can be seen from Figure (a), when the diameter of the suction anchor is constant, with the increase of the length-diameter ratio L / D, the uplift force of the suction anchor also continuously increases, and when v < 25 mm / s, the slope of the curve also continuously increases.

[0104] Figure 6 Figure (b) shows the relationship between the uplift force and the uplift rate when the length of the suction anchor is fixed at L = 50 and the length-diameter ratios L / D are 0.83, 1, and 1.25 respectively. Figure 6 As can be seen from Figure (b), when the length of the suction anchor is constant, with the increase of the length-diameter ratio L / D, the uplift force of the suction anchor continuously decreases;

[0105] From Figure 6 Figure (a) and Figure 6 Figure (b), it can be seen that the uplift force of the suction anchor first increases and then basically remains unchanged with the increase of the loading rate, and the relationship between the two is a non-linear relationship. When the uplift rate v of the suction anchor < 25 mm / s, the growth rate of the curve is relatively fast; when the uplift rate 25 mm / s < v < 100 mm / s, the growth rate of the curve decreases; when the uplift rate v > 100 mm / s, the curve basically remains unchanged. This is because with the increase of the loading rate, the failure mode of the suction anchor foundation has changed.

[0106] The above is only the preferred embodiment of the present invention and does not impose any limitation on the present invention. Any person skilled in the art within the technical field, without departing from the technical solution of the present invention, makes any form of equivalent substitution or modification and other changes to the technical solution and technical content disclosed by the present invention, all of which belong to the content not departing from the technical solution of the present invention and still fall within the protection scope of the present invention.

Claims

1. A method for predicting pullout capacity of a suction caisson, characterized by, It comprises the following steps: S1: Establishing a suction anchor pulling visualization experimental device; S2: Using the experimental device, carrying out suction anchor pulling experiments under different working conditions, and obtaining the relationship curve of pulling force and displacement under different working conditions through data processing; S3: According to the relationship curve of pulling force and displacement under different working conditions, generating a "factor-pulling force" database of the suction bucket; the factors include the length of the bucket of the circular bucket suction anchor, the diameter of the bucket of the circular bucket suction anchor, and the pulling speed; S4: Selecting a BP-RF machine learning algorithm for small sample data prediction, specifically: S4.1: Establish the input-output relationship table, set different factors as , constitute the input matrix X; S4.2: mapping the input-output relationship table to a BP neural network model, learning a mapping through training , predicting an output target; S4.3: mapping the input-output relationship table to a random forest model, learning , the prediction output target; S4.4: Weighted fusion of BP neural network model prediction results and random forest model prediction results to obtain the final prediction results; S5: Comparing and analyzing the prediction results and the measured results, and judging the influence law of the pulling force.

2. The suction bucket pullout capacity prediction method of claim 1, wherein, In step S1, the suction anchor pulling visualization experimental device comprises a suction anchor system, a counterforce frame, and a control mechanism: The suction anchor system comprises a hollow circular bucket suction anchor with a closed top end and an open bottom end, a pull rod connected at the center of the top end of the hollow circular bucket, and a driving mechanism connected to the pull rod; a tension sensor is installed on the pull rod, and the tension sensor is connected to a tension sensor data acquisition instrument; a drainage and exhaust hole with a valve is also provided on the anchor cover; The hollow circular bucket is fixed in a glass box filled with sand, and the glass box is located in the counterforce frame; the control mechanism is signal connected with the suction anchor system.

3. The suction bucket pullout capacity prediction method of claim 2, wherein, The sand is transparent soil; the counterforce frame is welded from square steel pipes; the driving mechanism is a linear servo electric cylinder for providing different speeds to drive the pull rod to move, with a lifting range of 0-300mm, a pulling force range of 0-1000N, and a moving speed range of 1-500mm / s; the control mechanism is a computer.

4. The suction bucket uplift capacity prediction method of claim 2, wherein, Step S2 specifically comprises: S2.1: Designing working conditions, the parameters that can be changed including the inner diameter of the suction bucket, the outer diameter of the suction bucket, the length of the suction bucket, and the pulling force loading rate; S2.2: Filling the configured sand into the organic glass box; S2.3: Using the driving mechanism to penetrate the suction anchor into the sand in the glass box; S2.4: After the suction anchor is penetrated into the sand, it is left for a period of time; S2.5: Pulling the suction anchor upward and recording the readings of the tension sensor; S2.6: Using the control mechanism to process the data, and further converting the tension sensor curve of the tension sensor changing with time into a relationship curve of the pulling force changing with displacement.

5. The suction bucket uplift capacity prediction method of claim 4, wherein, In step S2.5, the pulling method of the suction anchor is displacement control, and the pulling speed is controlled by the driving mechanism to pull the suction anchor at different loading rates.

6. The suction bucket pullout capacity prediction method of claim 1, wherein, In step S4.4, the weighted fusion formula is: ； In the formula, : the mixed weight, reflecting the contribution of two models, adjusting The value of can be found by the validation set to find the best value; is the prediction result of BP neural network; is the prediction result of random forest.

7. The suction bucket pullout capacity prediction method of claim 1, wherein, In step S5, the prediction results and the measured results are compared to analyze the influence trend of different factors on the pulling force, and the influence law of the pulling force is comprehensively judged by image analysis and machine learning methods.

Images