Digital twin method for real-time prediction of bearing performance of screw piles

Abstract

The present invention discloses a digital twin method for real-time prediction of the bearing performance of spiral piles, which belongs to the field of intelligent construction engineering. The method comprises: creating a user interface for inputting parameters based on UE, and creating a blueprint for automatically generating a corresponding spiral pile model according to the parameters. Based on the spiral pile sensor protocol, the sensor data is transmitted to the UE, and the virtual spiral pile in the UE is driven in real time. Spiral piles with different size parameters are created, and mechanical analysis is performed to obtain their respective load-settlement data and curves. Based on a convolutional neural network, the load-settlement data is trained to obtain a spiral pile load-settlement prediction agent model. Based on the user input parameters in the UE, the input is transmitted to the agent model in Python via TCP communication. Based on the agent model, the corresponding settlement and ultimate bearing capacity are predicted, and transmitted back to the UE for visualization. This method solves the problems of low efficiency and many limitations in the traditional prediction of the ultimate bearing capacity of spiral piles.

Description

Technical Field

[0001] The present invention relates to a digital twin method for real-time prediction of the bearing performance of a screw pile, and belongs to the field of intelligent construction engineering. Background Art

[0002] The bearing performance of a spiral pile includes the maximum load it can withstand. The bearing capacity directly affects the scope of use and bearing capacity of the spiral pile in engineering projects. Therefore, the prediction of the ultimate bearing capacity is very important. It directly affects the safety, economy, and efficiency of engineering projects, and is an indispensable key link in the engineering design and implementation process. At present, the main methods for predicting the ultimate bearing capacity of spiral piles include the single-disc bearing capacity calculation method, the cylindrical shear method, and the torque experience method. However, these methods have their own applicable conditions, and different methods must be used for different spiral piles. Moreover, these methods either require complex calculations and a lot of computational effort, or need to be used at the actual construction site. There are many limitations and low efficiency in predicting the ultimate bearing capacity of spiral piles. Summary of the Invention

[0003] The purpose of the present invention is to overcome the inefficiency and many limitations of the existing technology, and to provide a digital twin method that can quickly, accurately and unrestrictedly predict the bearing performance of screw piles in real time, so as to solve the problems of low efficiency and many limitations of traditional screw pile ultimate bearing capacity prediction.

[0004] The technical solution adopted by the present invention is:

[0005] A digital twin method for real-time prediction of bearing performance of screw piles includes the following steps:

[0006] Step 1: Create a user interface for inputting parameters based on UE. The parameters input in the user interface include the length of the screw pile, the outer diameter of the screw pile, the inner diameter of the screw pile, the number of spiral blades, the blade diameter, and the blade spacing. Automatically generate a screw pile model with the corresponding size and appearance based on the parameters input in the user interface, and create a blueprint Actor for the screw pile model.

[0007] Step 2: driving the screw pile model in real time based on sensor data; wherein the sensor data includes axial force, torque, rotational speed, and linear speed of the screw pile;

[0008] Step 3: Create multiple screw piles with different size parameters. Based on finite element analysis, create a soil model, define material properties, establish an assembly, define contact and interaction, divide the mesh, define boundary conditions and loads, set the analysis steps, submit the analysis, post-process the results, extract the result data, obtain the load and corresponding settlement values ​​at the pile top, and import them into the spreadsheet module. By inserting a broken line chart, obtain the load-settlement curve of each pile top. Based on the load-settlement curve of each pile top, obtain the ultimate bearing capacity of each pile top;

[0009] Step 4: Based on the convolutional neural network model, the various dimensions, load-settlement data, and ultimate bearing capacity of the screw pile are divided and trained as a data set. During the division and training process, the data set is divided into a training set, a validation set, and a test set. The feature data is scaled through normalization so that all features are within the same numerical range. The hierarchical structure of the convolutional neural network model is defined, using a combination of convolutional layers, pooling layers, flattening layers, and fully connected layers. The ReLU activation function and mean square error loss function are set. The data from the training set are input into the convolutional neural network model for training, and the parameters are adjusted to minimize the loss function. After training, the performance of the final model is evaluated using the data from the test set. Various indicators are calculated to quantify the prediction effect. After the evaluation is completed, the model is saved to obtain the screw pile load-settlement prediction agent model.

[0010] Step 5: Input parameters based on the user interface in the UE, add an input parameter to store the applied load, save all input parameters, and transmit them to the agent model in Python through TCP communication;

[0011] Step 6: The applied load and screw pile size are passed as input to the screw pile load-settlement prediction agent model, and the corresponding settlement and ultimate bearing capacity are predicted. The predictions are then transmitted back to the UE. In the user interface, the settlement and ultimate bearing capacity values ​​are visualized using the blueprint nodes EventTick and Set Variable. In the blueprint Actor of the screw pile model, the blueprint nodes Get Actor Location, Set Actor Location, Lerp, Timeline, and Branch are created and connected to adjust the settlement position of the screw pile model.

[0012] Furthermore, in step 1, a screw pile model of corresponding size and appearance is automatically generated according to the parameters input in the user interface, and a blueprint Actor of the screw pile model is created. By adding a static mesh component, instantiating the static mesh component, and a spline component, creating and connecting the blueprint nodes Set Relative Scale 3D, Add Instance, For Loop, and obtaining the transformation at the distance along the spline, a blueprint Actor for automatically generating a screw pile model of corresponding size and appearance according to the parameters is obtained.

[0013] Furthermore, the step 2 includes:

[0014] The screw pile model generated in step 1 is data-driven and synchronized with the on-site screw piles. Based on the screw pile sensor protocol, relevant code is written to transmit the sensor data to the UE in real time. By adding a rotation movement component to the blueprint Actor of the screw pile model, create and connect the blueprint nodes Set All Physics Linear Velocity and SetRotation Rate to achieve real-time synchronization of the motion status of the virtual screw pile and the on-site screw pile.

[0015] Furthermore, the step 3 includes:

[0016] Create multiple virtual screw pile models with different size parameters and import them into HyperMesh software for meshing;

[0017] After meshing, the data was imported into Abaqus for static load test mechanical analysis and soil model creation. The material properties of soil and screw piles were defined.

[0018] Build the assembly, combining the soil and screw pile components, ensuring good contact between the bottom of the pile and the soil;

[0019] Define the frictional contact and interaction between soil and screw piles;

[0020] Mesh the soil; define boundary conditions and loads, fix the soil around the sides and bottom, and apply a gradually increasing load to the top of the screw pile;

[0021] Set up analysis steps and submit analysis;

[0022] After the results are processed, the load at the top of the pile and the corresponding settlement values ​​are obtained using the result data extraction function;

[0023] Import it into Excel and insert a line chart to obtain the respective load-settlement data and curves. According to the curve, the load at which the pile head settlement is equal to 5% of the spiral blade diameter is obtained as the ultimate bearing capacity of the current spiral pile.

[0024] Furthermore, the step 4 includes:

[0025] The load-settlement data and ultimate bearing capacity data of the screw piles with different size parameters obtained in step 3 are used as a data set for analysis and training;

[0026] Based on the convolutional neural network (CNN) in machine learning, the various dimensions of the screw pile, including the length, outer diameter, inner diameter, number of spiral blades, blade diameter, blade spacing, and load, are used as input, and the settlement and ultimate bearing capacity are used as output to form a data set for division and training.

[0027] The dataset is divided into three parts: training set, validation set, and test set;

[0028] Normalization is used to scale feature data so that all features are within the same numerical range, which speeds up model convergence and improves prediction performance.

[0029] Define the model hierarchy, using a combination of convolutional layers, pooling layers, flattening layers, and fully connected layers;

[0030] Set the ReLU activation function to improve the nonlinear fitting ability of the model, and set the mean square error loss function to measure the accuracy of the prediction;

[0031] Input the training set data into the model for training, adjust the learning rate and hyperparameters as appropriate, and optimize the model architecture or training process to minimize the loss function. Methods for optimizing performance include adjusting the number of layers, the number of nodes, and regularization.

[0032] After training, the test set data is used to evaluate the performance of the final model, and the mean square error, root mean square error, and mean absolute error are calculated to quantify the prediction effect;

[0033] After the evaluation is completed, the model is saved to obtain the screw pile load-settlement prediction proxy model.

[0034] Furthermore, the step 5 includes:

[0035] Input parameters based on the user interface in the UE, add an input parameter to store the applied load, save all input parameters, and add all input parameters one by one to an array;

[0036] The array is transmitted to Python via TCP communication, the array is processed according to the sequence number of the array to obtain the value corresponding to each input parameter, and the value corresponding to each input parameter is input into the screw pile load-settlement prediction agent model.

[0037] Furthermore, in step 6, the data processed in step 5 is used as input, and a prediction is performed based on the spiral pile load-settlement prediction agent model to predict the ultimate bearing capacity of the spiral pile of the corresponding size and the settlement corresponding to the load, which are then transmitted back to the UE for visualization. The settlement and ultimate bearing capacity values ​​are visualized in the user interface using the blueprint nodes EventTick and SetVariable. A vector type variable is added to the spiral pile blueprint, and the blueprint nodes Get Relative Location, Set Relative Location, Lerp, Timeline, and Branch are created and connected to obtain the initial position of the virtual spiral pile and store it in the vector type variable. The new position vector is then obtained through the incoming settlement value, and the blueprint nodes Add Custom Event and Move ComponentTo are created and connected to control the virtual spiral pile to adjust its position to the corresponding settlement position.

[0038] The digital twin method for real-time prediction of the bearing capacity of screw piles provided by the present invention has the following beneficial effects:

[0039] The present invention can not only improve the efficiency of the prediction of the ultimate bearing capacity of the entire screw pile, reduce the workload in predicting the ultimate bearing capacity of the screw pile, and is applicable to most situations, times and places, but also ensure the safety and reliability of the engineering structure, and also help to optimize the design and selection of the screw pile, avoid over-design or unnecessary cost expenditure, and improve the economy and construction efficiency of the project. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the application.

[0041] Figure 1 The figure is a flowchart of a digital twin method for real-time prediction of the bearing performance of a screw pile, showing an exemplary embodiment.

[0042] The above drawings illustrate specific embodiments of the present invention, which will be described in more detail below. These drawings and the accompanying description are not intended to limit the scope of the present invention in any way, but rather to illustrate the concept of the present invention to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0043] The technical solutions of the present invention will be described clearly and completely below with reference to examples. Obviously, the embodiments described are only a part of the embodiments of the present invention, rather than all of the embodiments. The technical solutions and technical features provided in each part of the present invention, including the following description, can be combined with each other unless there is a conflict. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0044] The embodiment of the present invention provides a digital twin method for real-time prediction of the bearing performance of screw piles, such as Figure 1 As shown, the method includes the following steps S101 to S106.

[0045] S101. Create a user interface based on UE for inputting parameters, including the length of the screw pile, the outer diameter of the screw pile, the inner diameter of the screw pile, the number of spiral blades, the blade diameter, and the blade spacing. By adding static mesh components, instantiating static mesh components, and spline components, create and connect blueprint nodes such as "Set Relative Scale 3D", "Add Instance", "ForLoop", and "Get Transformation at Distance Along Spline" to automatically generate a blueprint Actor for the screw pile model of the corresponding size and appearance according to the parameters.

[0046] S102. The generated virtual spiral pile is data-driven and synchronized with the on-site spiral pile. Based on the spiral pile sensor protocol, relevant code is written to transmit sensor data in real time to the UE. The data includes the axial force, torque, rotation speed, and linear speed of the spiral pile. By adding a rotation movement component to the spiral pile blueprint, creating and connecting blueprint nodes such as "SetAll PhysicsLinearVelocity" and "Set Rotation Rate", the motion status of the virtual spiral pile and the on-site spiral pile are synchronized in real time.

[0047] S103. Create multiple virtual screw pile models with different size parameters, import them into HyperMesh software for meshing, and then import them into Abaqus for static load test mechanical analysis. Create a soil model; define the material properties of the soil and screw pile; establish an assembly to combine the soil and screw pile components to ensure good contact between the bottom of the pile and the soil; define the friction contact and interaction between the soil and the screw pile; mesh the soil; define boundary conditions and loads, fix the soil around the sides and bottom, and apply a gradually increasing load to the top of the screw pile; set the analysis steps and submit the analysis; post-process the results, use the result data extraction function to obtain the load and corresponding settlement values ​​at the pile top, import them into Excel, and insert a broken line chart to obtain the respective load-settlement data and curves. Then, obtain the curve fitting formula through fitting. According to the curve fitting formula, the load at which the pile head settlement is equal to 5% of the screw blade diameter is obtained as the ultimate bearing capacity of this screw pile.

[0048] S104. The obtained load-settlement data and ultimate bearing capacity data for screw piles of various dimensional parameters are used as a dataset for analysis and training. Based on the convolutional neural network (CNN) in machine learning, the various screw pile dimensions (including screw pile length, screw pile outer diameter, screw pile inner diameter, number of screw blades, blade diameter, blade spacing, and load) are used as inputs, and the settlement and ultimate bearing capacity are used as outputs to form a dataset for partitioning and training. The total dataset is divided into three parts: a training set, a validation set, and a test set. The feature data is scaled through normalization to ensure that all features are within the same numerical range, accelerating model convergence and improving prediction performance. A model hierarchy is defined, using a combination of convolutional layers, pooling layers, flattening layers, and fully connected layers. A ReLU activation function is set to enhance the model's nonlinear fitting capabilities, and a mean squared error loss function is set to measure prediction accuracy. The training set data is input into the model for training. The learning rate is adjusted as appropriate, and hyperparameters are adjusted to optimize the model architecture or training process, including adjusting the number of layers, nodes, and regularization to minimize the loss function. After training, the test data was used to evaluate the performance of the final model. Metrics such as mean square error, root mean square error, and mean absolute error were calculated to quantify the prediction effect. After the evaluation was completed, the model was saved to obtain the proxy model for the prediction of screw pile load and settlement.

[0049] S105. Input parameters based on the user interface in the UE, and add and store an input parameter for applying a load. Enter the pile length of 7 meters, the number of blades of 1, the blade diameter of 456 mm, the pile outer diameter of 342 mm, the pile inner diameter of 242 mm, the blade distance from the pile bottom of 50 cm, and the load of 100 kN in sequence. Save all the input parameters, add them one by one to an array, transmit this array to Python via TCP communication, process the array according to the array sequence number, obtain the value corresponding to each input parameter, and put it into the screw pile load-settlement prediction agent model after processing.

[0050] S106. The processed data is used as input and a prediction is performed based on the screw pile load-settlement prediction agent model. The ultimate bearing capacity of the screw pile of the corresponding size and the settlement corresponding to the load are predicted to be 160 kN and 60 mm. The data is then transmitted back to the UE for visualization. In the user interface, the settlement and ultimate bearing capacity values ​​are visualized using blueprint nodes such as "Event Tick" and "Set Variable." That is, "Ultimate bearing capacity: 160 kN" is displayed on the screen. A vector variable is added to the screw pile blueprint, and then blueprint nodes such as "Get Relative Location," "Set Relative Location," "Lerp," "Timeline," and "Branch" are created and connected to obtain the initial position of the virtual screw pile and store it in this vector variable. The new position vector is then obtained based on the incoming settlement value. Blueprint nodes such as "Add Custom Event" and "Move Component To" are created and connected to control the virtual screw pile to adjust its position to the corresponding settlement position, that is, move the screw pile downward by 60 mm.

[0051] Other embodiments of the present invention will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the claims.

[0052] It should be understood that the present invention is not limited to the exact construction described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof, which is limited only by the appended claims.

Claims

1. A digital twin method for real-time prediction of the bearing capacity of screw piles, characterized by: The following steps are involved: Step 1: Create a user interface for inputting parameters based on UE. The parameters input in the user interface include the length of the screw pile, the outer diameter of the screw pile, the inner diameter of the screw pile, the number of spiral blades, the blade diameter, and the blade spacing. Automatically generate a screw pile model with the corresponding size and appearance based on the parameters input in the user interface, and create a blueprint Actor for the screw pile model. Step 2: driving the screw pile model in real time based on sensor data; wherein the sensor data includes axial force, torque, rotational speed, and linear speed of the screw pile; Step 3: Create multiple screw pile models with different size parameters. Based on finite element analysis, create a soil model, define material properties, establish an assembly, define contact and interaction, divide the mesh, define boundary conditions and loads, set the analysis steps, submit the analysis, post-process the results, extract the result data, obtain the load and corresponding settlement values ​​at the pile top, and import them into the spreadsheet module. By inserting a broken line chart, obtain the load-settlement curve of each pile top. Based on the load-settlement curve of each pile top, obtain the ultimate bearing capacity of each pile top; Step 4: Based on the convolutional neural network model, the various dimensions, load-settlement data, and ultimate bearing capacity of the screw pile are divided and trained as a data set. During the division and training process, the data set is divided into a training set, a validation set, and a test set. The feature data is scaled through normalization so that all features are within the same numerical range. The hierarchical structure of the convolutional neural network model is defined, using a combination of convolutional layers, pooling layers, flattening layers, and fully connected layers. The ReLU activation function and mean square error loss function are set. The data from the training set are input into the convolutional neural network model for training, and the parameters are adjusted to minimize the loss function. After training, the performance of the final model is evaluated using the data from the test set. Various indicators are calculated to quantify the prediction effect. After the evaluation is completed, the model is saved to obtain the screw pile load-settlement prediction agent model. Step 5: Input parameters based on the user interface in the UE, add an input parameter to store the applied load, save all input parameters, and transmit them to the agent model in Python through TCP communication; Step 6: Pass the applied load and screw pile size as input to the screw pile load-settlement prediction agent model, predict the corresponding settlement and ultimate bearing capacity, and then transmit it back to the UE. Use the blueprint nodes EventTick and SetVariable in the user interface to visualize the settlement and ultimate bearing capacity values. Create and connect the blueprint nodes GetActor Location, SetActor Location, Lerp, Timeline, and Branch in the blueprint Actor of the screw pile model to adjust the settlement position of the screw pile model.

2. The digital twin method for real-time prediction of bearing capacity of screw piles according to claim 1, characterized in that: In the step 1, a screw pile model of corresponding size and appearance is automatically generated according to the parameters input in the user interface, and a blueprint Actor of the screw pile model is created. By adding a static mesh component, instantiating the static mesh component, and a spline component, creating and connecting the blueprint nodes Set Relative Scale 3D, Add Instance, For Loop, and obtaining the transformation at the distance along the spline, a blueprint Actor for automatically generating a screw pile model of corresponding size and appearance according to the parameters is obtained.

3. The digital twin method for real-time prediction of bearing capacity of screw piles according to claim 1, characterized in that: The step 2 includes: The screw pile model generated in step 1 is data-driven and synchronized with the on-site screw piles. Based on the screw pile sensor protocol, relevant code is written to transmit the sensor data to the UE in real time. By adding a rotation movement component to the blueprint Actor of the screw pile model, creating and connecting the blueprint nodes SetAll Physics LinearVelocity and Set RotationRate, the motion status of the screw pile model and the on-site screw piles can be synchronized in real time.

4. The digital twin method for real-time prediction of bearing capacity of screw piles according to claim 1, characterized in that: The step 3 includes: Create multiple screw pile models with different size parameters and import them into HyperMesh software for meshing; After meshing, the data was imported into Abaqus for static load test mechanical analysis and soil model creation. The material properties of soil and screw piles were defined. Build the assembly, combining the soil and screw pile components, ensuring good contact between the bottom of the pile and the soil; Define the frictional contact and interaction between soil and screw piles; Mesh the soil; define boundary conditions and loads, fix the soil around the sides and bottom, and apply a gradually increasing load to the top of the screw pile; Set up analysis steps and submit analysis; After the results are processed, the load at the top of the pile and the corresponding settlement values ​​are obtained using the result data extraction function; Import it into Excel and insert a line chart to obtain the respective load-settlement data and curves. According to the curve, the load corresponding to the target settlement is obtained as the ultimate bearing capacity of the current screw pile; wherein, the target settlement is the point where the pile head settlement is equal to 5% of the spiral blade diameter.

5. The digital twin method for real-time prediction of bearing capacity of screw piles according to claim 1, characterized in that: The step 4 comprises: The load-settlement data and ultimate bearing capacity data of the screw piles with different size parameters obtained in step 3 are used as a data set for analysis and training; Based on the convolutional neural network (CNN) in machine learning, the various dimensions of the screw pile, including the length, outer diameter, inner diameter, number of spiral blades, blade diameter, blade spacing, and load, are used as input, and the settlement and ultimate bearing capacity are used as output to form a data set for division and training. The dataset is divided into three parts: training set, validation set, and test set; Normalization is used to scale feature data so that all features are within the same numerical range, which speeds up model convergence and improves prediction performance. Define the model hierarchy, using a combination of convolutional layers, pooling layers, flattening layers, and fully connected layers; Set the ReLU activation function to improve the nonlinear fitting ability of the model, and set the mean square error loss function to measure the accuracy of the prediction; Input the training set data into the model for training, adjust the learning rate and hyperparameters as appropriate, and optimize the model architecture or training process to minimize the loss function. Methods for optimizing performance include adjusting the number of layers, the number of nodes, and regularization. After training, the test set data is used to evaluate the performance of the final model, and the mean square error, root mean square error, and mean absolute error are calculated to quantify the prediction effect; After the evaluation is completed, the model is saved to obtain the screw pile load-settlement prediction proxy model.

6. The digital twin method for real-time prediction of bearing capacity of screw piles according to claim 1 is characterized in that: The step 5 comprises: Input parameters based on the user interface in the UE, add an input parameter to store the applied load, save all input parameters, and add all input parameters one by one to an array; The array is transmitted to Python via TCP communication, the array is processed according to the sequence number of the array to obtain the value corresponding to each input parameter, and the value corresponding to each input parameter is input into the screw pile load-settlement prediction agent model.

7. The digital twin method for real-time prediction of bearing capacity of screw piles according to claim 1 is characterized in that: In step 6, the data processed in step 5 is used as input, and a prediction is performed based on the spiral pile load-settlement prediction agent model to predict the ultimate bearing capacity of the spiral pile of the corresponding size and the settlement corresponding to the load, which are then transmitted back to the UE for visualization. The settlement and ultimate bearing capacity values ​​are visualized in the user interface using the blueprint nodes Event Tick and SetVariable. A vector type variable is added to the spiral pile blueprint, and the blueprint nodes Get RelativeLocation, Set Relative Location, Lerp, Timeline, and Branch are created and connected to obtain the initial position of the spiral pile model and store it in the vector type variable. The new position vector is then obtained using the incoming settlement value, and the blueprint nodes Add Custom Event and Move Component To are created and connected to control the spiral pile model to adjust its position to the corresponding settlement position.

Images