A method for predicting overall horizontal force of deepwater jacket based on physical characteristics

Abstract

A physical feature-based overall horizontal force prediction method for deepwater jacket platforms belongs to the field of digital twinning of deepwater jackets. The method is based on the physical calculation formula of the overall horizontal force of the platform, and combines the convolutional neural network to capture the nonlinear mapping relationship between the wind, wave and current environmental loads and the overall horizontal force of the platform, and establishes an intelligent prediction method for the overall horizontal force of the deepwater jacket platform based on physical features. The invention can accurately predict the overall horizontal force of the deepwater jacket platform under the action of wind, wave, current and other marine environmental loads, so as to realize the real-time monitoring and performance evaluation of the overall performance state of the platform structure, and can be widely applied in the field of digital twinning of deepwater jackets.

Description

Technical Field

[0001] This invention relates to the field of marine engineering and employs a method for predicting the overall horizontal force of a deep-water jacket platform based on a convolutional neural network (CNN). Background Technology

[0002] With the continuous development of marine resource development and marine engineering, deep-water jacket platforms, as important facilities for offshore oil and gas extraction, are crucial for ensuring a stable supply of marine resources and the safety of personnel. Under the influence of complex marine environmental loads such as wind, waves, and currents, deep-water jacket platforms exhibit significant nonlinear dynamic characteristics, posing a major challenge to predicting the overall performance of the platform.

[0003] Traditional deepwater jacket platform design and performance evaluation methods often rely on empirical formulas or simplified physical models. These methods have limitations in handling nonlinear dynamic problems and struggle to accurately capture the platform's response characteristics under complex marine environmental loads. Furthermore, while traditional numerical simulation methods can provide relatively accurate simulation results, their high computational cost makes them unsuitable for real-time monitoring and rapid response. Summary of the Invention

[0004] To address the problems existing in current technologies, this invention provides a method for predicting the overall horizontal force of deep-water jacket platforms based on physical characteristics. It employs a convolutional neural network (CNN) to predict the overall horizontal force of deep-water jacket platforms, thus overcoming the insufficient accuracy and low efficiency of existing technologies. This invention constructs a high-precision overall index prediction model to accurately predict the overall horizontal force of deep-water jacket platforms under marine environmental loads such as wind, waves, and currents. This model improves the accuracy and efficiency of prediction, providing technical support for the design, monitoring, and maintenance of deep-water jacket platforms.

[0005] The technical solution adopted in this invention is as follows: a method for predicting the overall horizontal force of a deep-water jacket based on physical characteristics: the prediction model for the overall horizontal force of a deep-water jacket includes a platform overall horizontal force feature dataset module, and a method for predicting the overall horizontal force of a deep-water jacket based on a convolutional neural network (CNN); The prediction method specifically includes the following steps: S1. Based on the structural characteristics of the deep-water jacket platform, extract monitoring data and preprocess it. S2. Combining the solution method and formula of the horizontal force of the deep-water jacket, the horizontal force of the jacket is calculated using its monitoring data; S3. Establish the mapping relationship between monitoring data in deep-water jacket structures and the overall horizontal force of the platform; S4. Utilize a physical feature-based intelligent prediction algorithm for the overall horizontal force of the deep-water jacket platform to predict the overall horizontal force; this includes the following sub-steps: S4.1 Construct a dataset of the overall performance characteristics of the platform into a database, and divide the total dataset into a training set, a validation set, and a test set; S4.2 uses the combined loads of wind, waves, and currents in the marine environment as model inputs and the overall horizontal force of the deep-water jacket as model outputs; S4.3 Construct a CNN prediction model structure that includes an input layer, convolutional layer, Dropout layer, fully connected layer, and output layer; S4.4 Combines the prediction model and the database to iteratively train the CNN prediction model, including parameter initialization, backpropagation, parameter update, and cross-validation. S4.5 The above steps are used to evaluate and optimize the overall horizontal force prediction model of the deep-water jacket platform to obtain the optimal hyperparameters of the model. Furthermore, step 2 specifically includes the following sub-steps: S2.1 The axial force of the bracing and support legs at the measuring points is calculated using stress-strain sensors installed on the diagonal bracing and support legs at the base of the deep-water jacket foundation. F X and F T ; S2.2 Establish the overall force equilibrium equations for the deep-water jacket structure and describe the environmental loads. F and the support reactions at fulcrums A and B F A , F B Static equilibrium relationship; S2.3 Establish the local force equilibrium equations for the deep-water jacket structure, describing the support reactions at supports A and B. F A , F B Axial force of related diagonal braces F XA 、F XB ， Axial force of supporting leg F TA , F TB Static equilibrium relationship;

[0006] S2.4 Solving for environmental loads F And the overall level of the platform was calculated. F w ;

[0007] in: It is a unit vector in the horizontal direction.

[0008] Furthermore, in step S4.2, the load data includes wind speed, wind direction, flow speed, flow direction, wave direction, wave height, and period.

[0009] Advantages of this invention: 1. Compared with traditional deep-water jacket platform design and performance evaluation methods that often rely on empirical formulas or simplified physical models, this invention can better handle nonlinear problems and accurately capture the platform's response characteristics under complex marine environmental loads.

[0010] 2. Compared with traditional numerical simulation methods, the prediction method in this invention has lower computational cost and can meet the needs of real-time monitoring and rapid response more quickly.

[0011] 3. This invention provides a detailed explanation of the solution method for the key features of the overall horizontal force of the platform, ensuring the accuracy and reliability of the input data and laying the foundation for high-precision prediction of the model.

[0012] 4. This invention establishes a mapping relationship between multiple monitoring data and the overall horizontal force of the deep-water jacket platform, and considers the index of the overall horizontal force of the deep-water jacket platform under the combined load distribution. It can capture the complex information in the prediction network more comprehensively and accurately, and thus more effectively predict the overall horizontal force of the deep-water jacket platform.

[0013] 5. This invention, by constructing a high-precision CNN prediction model, can obtain the overall performance status of the platform structure, which is crucial for realizing a digital twin of the platform's overall performance. The CNN-based overall horizontal force prediction model for the jacket platform achieves a prediction accuracy of 97.8% for the overall horizontal force of the platform. This helps to more accurately monitor and evaluate the operational status of deep-water jacket platforms in complex marine environments, contributing to the digital and intelligent development of marine engineering, and possessing potential engineering application value and innovative significance. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating the construction process of the platform's overall horizontal force physical characteristic dataset.

[0015] Figure 2 This is a flowchart of the training and validation process for a CNN-based overall horizontal force prediction model for pipe scaffolds.

[0016] Figure 3 This is a fitting result of the CNN-based model for predicting the overall horizontal force of the duct stent. Detailed Implementation

[0017] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. However, it should be understood that the drawings are provided only for a better understanding of the present invention and should not be construed as limiting the present invention.

[0018] A method for predicting the overall horizontal force of a deep-water jacket based on physical characteristics includes the following steps: S1. Based on the structural characteristics of the deep-water jacket platform, extract monitoring data with stable data quality and preprocess it. S2. Calculate the axial force of the bracing and support legs near the measuring point using stress-strain sensors installed at the base of the deep-water jacket support, and then deduce the overall horizontal force of the platform. F w Specifically, it includes the following sub-steps: S2.1 Calculates the axial force of the bracing and support legs near the measuring point using stress-strain sensors installed on the bracing and support legs at the base of the deep-water jacket. F X and F T ; S2.2 Establish the overall force equilibrium equations for the deep-water jacket structure and describe the environmental loads. F and the support reactions at fulcrums A and B F A , F B Static equilibrium relationship; S2.3 Establish the local force equilibrium equations for the deep-water jacket structure, describing the support reactions at supports A and B. F A , F B Axial force of related diagonal braces and supporting legs F XA 、F TA , F XB 、F TB Static equilibrium relationship;

[0019] S2.4 Solving for environmental loads F And the overall level of the platform was calculated. F w ;

[0020] in: It is a unit vector in the horizontal direction.

[0021] S3. Establish the mapping relationship between monitoring data in deep-water jacket structures and the overall horizontal force of the platform; S4. Utilize a physical feature-based intelligent prediction algorithm for the overall horizontal force of the deep-water jacket platform to accurately predict the overall horizontal force. This step specifically includes the following sub-steps: S4.1 Construct a dataset of the overall performance characteristics of the platform into a database, and divide the total dataset into a training set, a validation set, and a test set; S4.2 uses the combined loads of wind, waves, and currents in the marine environment as model inputs and the overall horizontal force of the deep-water jacket as model outputs; S4.3 Construct a CNN prediction model structure that includes an input layer, convolutional layer, Dropout layer, fully connected layer, and output layer; S4.4 Combines the prediction model and the database to iteratively train the CNN prediction model, including steps such as parameter initialization, backpropagation, parameter update, and cross-validation. S4.5 The above steps are used to evaluate and optimize the overall horizontal force prediction model of the deep-water jacket platform to obtain the optimal hyperparameters of the model. S4.6 Using a trained overall horizontal force prediction model for deep-water jacket platforms, the overall horizontal force of the deep-water jacket platform under marine environmental loads such as wind, waves, and currents is evaluated and tested using MAE, RMSE, and MAPE evaluation indicators, proving the rationality of the designed model. Figure 1 This is a flowchart illustrating the construction of the physical characteristic dataset of the overall horizontal force of the platform. By building a high-precision overall index prediction model, accurate prediction of the overall horizontal force of the deep-water jacket platform under marine environmental loads such as wind, waves, and currents is achieved. In the implementation process, the collected data of the deep-water jacket platform under different marine environmental loads are first normalized to adapt to the input requirements of the CNN model. Next, the overall performance characteristic dataset of the platform, specifically the overall horizontal force, is extracted.

[0022] The CNN model architecture consists of an input layer, two convolutional layers, two dropout layers, one fully connected layer, and an output layer. The convolutional layers use one-dimensional convolutions with a kernel size of 1*3 to adapt to the scale of the data features. Within the convolutional layers, local features are extracted from the data by sliding the convolutional kernel across the input feature map. To improve the model's generalization ability, two dropout layers are added instead of pooling layers.

[0023] After training, the trained CNN model is used to predict data on the test set, and the predicted values ​​are compared with the actual values ​​to evaluate the model's prediction accuracy. Model performance is tested and evaluated using metrics such as MAE and RMSE. Based on the evaluation results, the model structure or hyperparameters are adjusted to improve prediction accuracy.

[0024] Example 1 S1. Combining the structural characteristics of the deep-water jacket platform with the monitoring data from sensors (wind, waves, currents, and axial force information of the rods, etc.), after analysis, sensors with stable data quality were selected for analysis and modeling, and their data were processed. S2. Combining the calculation methods and formulas for the horizontal force of the deep-water jacket support, and using the axial forces of the 8 diagonal braces and 4 legs at the base of Liuhua 11-1, the overall horizontal force F of the platform is calculated. w S3. Establish the mapping relationship between monitoring data (environmental load data such as wind speed, wind direction, current velocity, current direction, wave direction, wave height, and period) and the overall horizontal force of the platform in the deep-water jacket; obtain the overall horizontal force characteristic dataset of the platform, and divide it into a 70% training set, a validation set, and a 30% test set. S4. By constructing a CNN structure containing an input layer, convolutional layer, dropout layer, fully connected layer and output layer, a smart prediction model for the overall horizontal force of deep-water jackets based on physical features is established. S5. Combining the above network prediction model and calling the database, iteratively train the CNN network model, including steps such as parameter initialization, backpropagation, parameter update, and cross-validation. The platform sensor network mapping model training process is as follows: Figure 2 As shown; S6. Through the training process of the evaluation and tuning of the sensor network mapping model, obtain the optimal hyperparameters of the model. The number of input features is 7, the number of convolutional neural network layers is 5, the number of hidden layers is 128, the batch size is 64, the number of training rounds is 1000, the number of early stopping rounds is 20, the learning rate is 0.001, the optimizer is Adam, and the loss function is mean squared error.

[0025] Predictions were made using untrained test set data on a pre-trained CNN-based model for predicting the overall horizontal force of a duct stent, and the predicted values ​​were plotted as a fitted curve with the actual values. Figure 3 The fitting effect of the CNN prediction model on the test set data is demonstrated, including the fitting of the overall horizontal force Fx along the X direction and the horizontal force Fy along the Y direction of the platform, proving the reliability of the model. Validation results on the test set show that the CNN-based guide frame overall horizontal force prediction model achieves a prediction accuracy of 97.8% for the overall horizontal force of the platform.

[0026] The above embodiments are only used to illustrate the present invention. The structure of the deep-water jacket, environmental load, etc. can be changed. All equivalent transformations and improvements made on the basis of the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A method for predicting the overall horizontal force of a deep-water jacket based on physical characteristics, characterized in that: Includes the following steps: S1. Based on the structural characteristics of the deep-water jacket platform, extract monitoring data and preprocess it. S2. Combining the solution method and formula of the horizontal force of the deep-water jacket, the horizontal force of the jacket is calculated using its monitoring data; S3. Establish the mapping relationship between monitoring data in deep-water jacket structures and the overall horizontal force of the platform; S4. Utilize a physical feature-based intelligent prediction algorithm for the overall horizontal force of the deep-water jacket platform to predict the overall horizontal force; this includes the following sub-steps: S4.1 Construct a dataset of the overall performance characteristics of the platform into a database, and divide the total dataset into a training set, a validation set, and a test set; S4.2 uses the combined loads of wind, waves, and currents in the marine environment as model inputs and the overall horizontal force of the deep-water jacket as model outputs; S4.3 Construct a CNN prediction model structure that includes an input layer, convolutional layer, Dropout layer, fully connected layer, and output layer; S4.4 Combines the prediction model and the database to iteratively train the CNN prediction model; S4.5 The above steps are used to evaluate and optimize the overall horizontal force prediction model of the deep-water jacket platform to obtain the optimal hyperparameters of the model.

2. The method for predicting the overall horizontal force of a deep-water jacket based on physical characteristics according to claim 1, characterized in that: Step 2 specifically includes the following sub-steps: S2.1 Calculates the axial force of the bracing and support legs near the measuring point using stress-strain sensors installed on the bracing and support legs at the base of the deep-water jacket. F X and F T ; S2.2 Establish the overall force equilibrium equations for the deep-water jacket structure to describe the environmental loads. F and the support reactions at fulcrums A and B F A , F B Static equilibrium relationship; S2.3 Establish the local force equilibrium equations for the deep-water jacket structure, describing the support reactions at supports A and B. F A , F B Axial force of related diagonal braces F XA 、F XB ， Axial force of supporting leg F TA , F TB Static equilibrium relationship; S2.4 Solving for environmental loads F And the overall level of the platform was calculated. F w ; in: It is a unit vector in the horizontal direction.

3. The method for predicting the overall horizontal force of a deep-water jacket based on physical characteristics according to claim 1, characterized in that: In step S4.2, the load data includes wind speed, wind direction, flow speed, flow direction, wave direction, wave height, and period.

4. The method for predicting the overall horizontal force of a deep-water jacket based on physical characteristics according to claim 1, characterized in that: In step S4.4, the CNN prediction model is iteratively trained, including parameter initialization, backpropagation, parameter update, and cross-validation.

Images