A real-time correction and state prediction method for jacket platform based on reduced-order model

Abstract

This invention discloses a real-time correction and state prediction method for jacket platforms based on a reduced-order model, relating to the field of marine engineering structural health monitoring and safety early warning technology. The method includes: establishing a finite element model of the jacket platform; obtaining a sample matrix; constructing a reduced-order model; reducing the order of the reduced-order model until its computational accuracy meets a preset accuracy requirement, obtaining a global reduced-order basis matrix and a reduced-order model that meets the preset accuracy requirement; reconstructing the stiffness matrix and mass matrix using the global reduced-order basis matrix; obtaining the corrected finite element model; inverting marine environmental loads based on on-site platform structural strain response data, deriving the marine environmental loads; inputting the marine environmental loads into a pre-trained LSTM model, and outputting time-series data containing future loads through the LSTM model; predicting the future operating state trend of the jacket platform. This invention improves the operational safety and reliability of jacket platforms.

Description

Technical Field

[0001] This application relates to the field of marine engineering structure health monitoring and safety early warning technology, and in particular to a method for real-time correction and state prediction of a jacket platform based on a reduced-order model. Background Technology

[0002] With the accelerated development of global marine energy, the development trend of offshore platform equipment is gradually showing characteristics of larger size, greater complexity, and deeper water. Currently, the total number of offshore platforms used for offshore oil and gas extraction worldwide exceeds 7,000, with jacket structures accounting for the vast majority. However, the design and operation of jacket platforms face multiple challenges, especially in complex marine environments. First, the structure of a jacket platform is extremely complex, consisting of multiple components such as jacket legs, support structures, and connection nodes. These components involve various materials and connection methods during design and construction, increasing the difficulty of structural analysis. Second, the platform is affected by environmental loads such as wind, waves, and currents during long-term operation. These loads are variable and uncertain, threatening the platform's structural performance and safety. Simultaneously, corrosion, fatigue, and wear in the marine environment can lead to gradual degradation of the platform's structural performance. This degradation requires timely detection and correction to prevent safety accidents. Traditional detection and maintenance methods rely on periodic inspections, which not only fail to detect instantaneous changes and potential faults in a timely manner but are also costly.

[0003] Finite element analysis (FEM) is one of the most commonly used methods in structural analysis. It divides complex structures into multiple small elements, each whose mechanical behavior can be described by a simple mathematical model. This method plays a crucial role in the design and evaluation of jacket platforms. However, traditional FEM methods suffer from high computational complexity and long processing times when dealing with complex structures, making real-time analysis and correction difficult. Furthermore, the establishment and verification of FEM models require substantial experimental data and computational resources, posing a significant challenge in practical applications. To address the high computational complexity of FEM, the reduced-order model technique has emerged. The reduced-order model extracts the main features of the structure, reducing the high-dimensional FEM model to a low-dimensional model, thereby significantly improving computational efficiency. Chinese invention patent application number 202410456210.5 discloses a digital twin monitoring method and system for gas pipelines based on a reduced-order model. By constructing a CAE simulation model, a reduced-order model, and a fault diagnosis model of the gas pipeline, combined with digital twin technology, it achieves real-time monitoring and fault diagnosis of the gas pipeline. Chinese invention patent application No. 202411748860.3 discloses a centrifugal impeller monitoring method and system based on a twin reduced-order model. By collecting blade tip vibration signals, preprocessing data, identifying dynamic parameters, and constructing and updating the reduced-order model, it realizes efficient, real-time and accurate monitoring of the operating status of large rotating machinery equipment, effectively solving the problems of large computational load and difficulty in extracting detuned parameters in traditional monitoring methods.

[0004] While model reduction techniques have made significant progress in improving computational efficiency, existing methods suffer from insufficient real-time correction capabilities when applied to jacket platform model modification. Current methods typically perform a single model correction in the initial stage, but due to the massive computational cost of this process, the model is not updated further during subsequent runs. As the platform operates over a long period, environmental changes and structural degradation gradually weaken the model's predictive power, rendering it unable to meet practical needs. Summary of the Invention

[0005] Therefore, it is necessary to provide a method for real-time correction and state prediction of the jacket platform based on a reduced-order model to address the above-mentioned technical problems.

[0006] The following technical solution is adopted in this specification:

[0007] This specification provides a method for real-time correction and state prediction of a jacket platform based on a reduced-order model, including:

[0008] Based on the initial structural parameters of the jacket platform, a finite element model of the jacket platform is established;

[0009] At least two load conditions are applied sequentially to the finite element model of the jacket platform, and the displacement field at different times under each load condition is extracted and combined into a sample matrix.

[0010] A reduced-order basis matrix is ​​constructed based on the sample matrix, and the sample matrix is ​​projected onto the reduced-order basis matrix to construct a reduced-order model;

[0011] The order of the reduced model is reduced. When the computational accuracy of the reduced model meets the preset accuracy requirements, the global reduced basis matrix and the reduced model that meets the preset accuracy requirements are obtained.

[0012] Reconstruct the stiffness and mass matrices using a global reduced-order basis matrix;

[0013] The numerical frequencies and modes of the jacket platform are calculated using the reconstructed stiffness matrix and mass matrix. The objective function is constructed by combining the acceleration response data of the jacket platform collected by the structural monitoring sensor. The correlation coefficients of the design variables of the objective function are adjusted by iterative optimization algorithm until the objective function converges, and the corrected finite element model is obtained.

[0014] Based on the corrected finite element model, a reduced-order finite element model is obtained, and time series data is applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time.

[0015] The beneficial effects of this technique are as follows: By constructing a reduced-order model that meets preset accuracy requirements, a cost-effective method for reducing the order of the finite element model of the jacket platform is provided. This allows for multiple reductions in the order of the finite element model, significantly reducing the computational load and increasing the computational speed. Furthermore, based on the reduced-order basis matrix, the mass and stiffness matrices of the platform structure are reconstructed. Iterative optimization algorithms are then used to calculate the objective function for finite element model correction based on real-time sensor data, enabling real-time correction of the finite element model and preventing weakening of its predictive ability due to environmental changes and structural degradation. Finally, the predicted environmental loads are applied to the corrected finite element model of the jacket platform, and combined with the model reduction technique, short-term predictions of the platform's future operating status trends are made, improving the accuracy of future operating status trend predictions for the jacket platform.

[0016] Optionally, before applying at least two load conditions sequentially to the finite element model of the jacket platform and extracting the displacement field at different times under each load condition, and combining the displacement fields into a sample matrix, the method further includes:

[0017] Based on the load variation characteristics of the historical load of the jacket platform in different time periods, the historical load is divided into at least two load conditions;

[0018] Constructing a reduced-order basis matrix based on the sample matrix specifically includes:

[0019] Construct a correlation matrix based on the sample matrix;

[0020] Construct the Lagrange function based on the correlation matrix and find its extreme value to obtain the eigenvalues ​​and eigenvectors of the correlation matrix. Sort the eigenvalues ​​in descending order and select the first N eigenvectors to construct a reduced-order basis matrix.

[0021] Based on the revised finite element model, a reduced-order finite element model is obtained, and time series data is applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time. Specifically, this includes:

[0022] The marine environmental load is obtained by inverting the strain response data of the on-site jacket platform structure; the marine environmental load is input into a pre-trained LSTM model, and the pre-trained LSTM model outputs time series data containing future loads.

[0023] The modified finite element model is reduced in order using a method that meets the preset accuracy requirements, resulting in a reduced-order finite element model. Time series data is then applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time.

[0024] Optionally, the sample matrix A = [a1, a2, ..., a...] i ], where a i For different times t j displacement field u(x) i ,t j The matrix assembled from these components. x i (i = 1, 2, ..., M) represents the i-th grid node, t j (j=1,2,…,N) represents the j-th time node of the finite element model of the jacket platform.

[0025] Optionally, the computational accuracy of the reduced-order model can be calculated as follows:

[0026]

[0027] Among them, R 2 N represents the computational accuracy of the reduced-order model. n This indicates the number of samples used to verify accuracy. This represents the i-th displacement field obtained based on the reduced-order model. This represents the i-th displacement field obtained based on the full-order model. N represents n The average value of each displacement field.

[0028] Stiffness matrix K POD for:

[0029]

[0030] Among them, P s s represents the number of stiffness matrices that make up the overall stiffness matrix. i G represents the number of modes involved in the stiffness matrix during the order reduction process. j (x i ) is the design variable x of the stiffness matrix i The relevant parts, Let C represent the stiffness matrix projected from the reduced basis matrix, and k represent the reduced basis matrix. is To design variables x with stiffness matrix i Irrelevant parts;

[0031] Mass matrix M POD for:

[0032]

[0033] Where, q i h represents the number of modes involved in the quality matrix during the order reduction process. j (y i ) Design variables y with the quality matrix i The relevant parts, Let m represent the mass matrix after projection through the reduced basis matrix. is Represents the design variable y with the quality matrix. i Irrelevant parts.

[0034] Optionally, the objective function is:

[0035]

[0036] Among them, f j f is the j-th order numerical frequency obtained from the reduced-order model. Ej Let be the j-th order numerical frequency obtained from indoor model experiments, r be the number of numerical modes matching the experimental mode, μ be the weighting coefficient, and MAC be the numerical frequency. j Let be the degree of similarity between the j-th order numerical mode and the experimental mode. Let θ be the j-th order mode obtained from the reduced-order model. j Let be the j-th mode obtained from the full-order model.

[0037] Optionally, future operating status trends include time series plots and predicted values ​​of displacement and strain; the method also includes:

[0038] The system displays future operational trends through a graphical interface. When these trends exceed safety thresholds, an alarm mode is activated, generating alarm information. This alarm information includes the time of the future fault, its specific location, the cause of the fault, and the corresponding handling measures.

[0039] Optionally, the future operational trend exceeds the safety threshold, specifically including:

[0040] When the first strut of the jacket platform yields, the overall displacement value is set as the first-level warning value. The displacement value of the first-level warning value is determined based on the overall displacement when the strut yields. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements.

[0041] When the first leg of the jacket platform yields, the overall displacement value is set as the secondary warning value. The displacement value of the secondary warning value is determined based on the overall displacement when the leg yields. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements.

[0042] When a node of the jacket platform fails, the overall displacement value is set as the level three warning value. The displacement value of the level three warning value is determined based on the overall displacement when the node fails. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements.

[0043] The alarm levels in the alarm mode include Level 1, Level 2, and Level 3.

[0044] This specification provides a real-time correction and state prediction system for a jacket platform based on a reduced-order model, including:

[0045] The finite element model building module is specifically used to build a finite element model of the jacket platform based on the initial structural parameters of the jacket platform.

[0046] The reduced-order model construction module is specifically used to apply at least two load conditions sequentially to the finite element model of the jacket platform, extract the displacement field at different times under each load condition, combine the displacement field into a sample matrix, construct a reduced-order basis matrix based on the sample matrix, project the sample matrix using the reduced-order basis matrix to construct a reduced-order model, reduce the order of the reduced-order model, and when the calculation accuracy of the reduced-order model meets the preset accuracy requirements, obtain the global reduced-order basis matrix and the reduced-order model that meets the preset accuracy requirements.

[0047] The model correction module is specifically used to reconstruct the stiffness matrix and mass matrix using the global reduced-order basis matrix; the numerical frequency and mode of the jacket platform are calculated using the reconstructed stiffness matrix and mass matrix; the objective function is constructed by combining the acceleration response data of the jacket platform collected by the structural monitoring sensor equipment; the correlation coefficient of the design variables of the objective function is adjusted by the iterative optimization algorithm until the objective function converges, and the corrected finite element model is obtained.

[0048] The state prediction module is specifically used to obtain a reduced-order finite element model based on the corrected finite element model, and apply time series data to the reduced-order finite element model to predict the future operating state trend of the jacket platform in real time.

[0049] This specification provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for real-time correction and state prediction of a jacket platform based on a reduced-order model.

[0050] This specification provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described method for real-time correction and state prediction of the jacket platform based on the reduced-order model.

[0051] The above-mentioned technical solutions adopted in this specification can achieve the following beneficial effects:

[0052] In the real-time correction and state prediction method for jacket platform based on a reduced-order model provided in this specification, the main modal features of the platform are extracted using POD technology to construct a reduced-order basis matrix. Based on this reduced-order basis matrix, the mass matrix and stiffness matrix of the platform structure are reconstructed. Then, using an iterative optimization algorithm, the objective function for finite element model correction is calculated based on real-time sensor data to achieve real-time correction of the finite element model. Based on the collected field strain response monitoring data, the load coefficient matrix is ​​accurately solved using numerical methods to further derive the marine environmental load on the overall model and predict the future trend of the load. The predicted environmental load is applied to the corrected finite element model of the jacket platform, and combined with the model reduction technology, the future operating state trend of the platform is predicted in the short term, improving the accuracy of the prediction of the future operating state trend of the jacket platform.

[0053] Furthermore, this invention, by combining model reduction techniques, real-time model correction methods, and machine learning algorithms, constructs a digital twin that is highly synchronized with the physical jacket platform. Based on real-time monitoring data and historical data, it can assess the platform's future health status and predict potential safety risks. This significantly improves the operational safety and reliability of the jacket platform, provides a scientific basis for platform health management in complex marine environments, and possesses significant application value and innovation. Attached Figure Description

[0054] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0055] Figure 1This document presents a flowchart illustrating a method for real-time correction and state prediction of a jacket platform based on a reduced-order model.

[0056] Figure 2 This document presents a schematic diagram of a real-time correction and state prediction system for a jacket platform based on a reduced-order model.

[0057] Figure 3 This specification provides a schematic diagram of a computer device for implementing a method for real-time correction and state prediction of a jacket platform based on a reduced-order model. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this specification clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in this specification without creative effort are within the scope of protection of this application.

[0059] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0060] Figure 1 This is a flowchart illustrating a real-time correction and state prediction method for a jacket platform based on a reduced-order model, as described in this specification. The method includes the following steps:

[0061] S101. Based on the initial structural parameters of the jacket platform, establish a finite element model of the jacket platform.

[0062] For example, based on the design drawings of the on-site jacket platform, initial structural parameters of the platform are collected, including material properties such as geometric dimensions, elastic modulus, density, and yield strength, as well as the connection methods between platform structures. Using these parameters, a finite element model of the jacket platform is constructed using finite element analysis software, ensuring the completeness and accuracy of the model.

[0063] The server mentioned in this manual can be a server set up on a business platform, or a device such as a desktop computer or laptop computer capable of executing the solution described in this manual. For ease of explanation, the following description will only focus on the server as the execution subject.

[0064] For example, (a) Design drawing preparation: collect complete design drawings of the jacket platform, including the three-dimensional structural drawings of the platform and a detailed material property table;

[0065] (b) Construction of 3D numerical model: Using SolidWorks 3D modeling software, a model with the same geometry as the design drawings is drawn;

[0066] (c) Finite Element Model Construction: Import the above three-dimensional numerical model into the finite element analysis software ANSYS to construct the finite element model of the jacket platform. The model should include all key components of the platform to ensure completeness and accuracy;

[0067] (d) Material Property Input: Input the platform's material properties, such as elastic modulus, density, and yield strength, into the ANSYS software. The material properties should be consistent with the data in the design drawings.

[0068] (e) Connection Method Definition: Define various connection methods between platform structures, such as welding, bolting, etc. Ensure that the parameters of each connection method conform to the actual jacket platform design.

[0069] In this embodiment, a variety of high-precision sensors are used, including accelerometers, strain sensors, anemometers, and radar wave meters, to ensure comprehensive data acquisition. The anemometers are installed on the pillars of the overhead crane's guardrail on the platform. The radar wave meters are installed on the deck of the platform near sea level. Structural monitoring sensors should be installed in key parts of the platform, such as support structures, connection nodes, and areas of maximum stress, to ensure the comprehensiveness and reliability of the data. The data collected by the sensors is transmitted to the central processing unit via wired or wireless communication.

[0070] S102. Apply at least two load conditions to the finite element model of the jacket platform in sequence, extract the displacement field at different times under each load condition, and combine the displacement fields into a sample matrix.

[0071] In this embodiment, before extracting the displacement field at different times under each load condition after applying at least two load conditions sequentially to the finite element model of the jacket platform and combining the displacement fields into a sample matrix, the method further includes:

[0072] Based on the load variation characteristics of the historical load on the jacket platform at different time periods, the historical load is divided into at least two load conditions.

[0073] Sample matrix A = [a1, a2, ..., a i ], where a i For different times t j displacement field u(x) i ,t j The matrix assembled from these components. x i (i = 1, 2, ..., M) represents the i-th grid node, t j (j=1,2,…,N) represents the j-th time node of the finite element model of the jacket platform.

[0074] For example, based on the historical loads of the jacket platform at the site, including parameters such as wind, waves, and currents, various load conditions are classified according to the load variation characteristics over different time periods. These load conditions are then sequentially applied to the finite element model of the jacket platform, and the time t under each load condition is recorded. j (j = 1, 2, ..., N) Grid node x i The displacement field u(x) at (i = 1, 2, ..., M) i ,t j ), and assemble them into matrix a. i The static displacement field under each load condition is assembled into a sample matrix A.

[0075] For example, (a) load condition determination: Based on the service environment of the on-site jacket platform, determine the environmental load it bears, mainly including wind, wave, and current loads. According to the load change characteristics of different time periods, the load can be divided into multiple working conditions.

[0076] (b) Load Application: Apply the corresponding load conditions in the finite element analysis software and perform dynamic analysis. Apply each load condition separately to ensure that the conditions applied each time are consistent with the actual environment;

[0077] (c) Displacement field recording: Record the model at different times t under each load condition. j displacement field u(x) i ,t j These displacement data will be used for subsequent construction of a reduced-order model;

[0078] (d) Displacement field matrix assembly: Assemble the displacement field data u(x) under each load condition. i ,t j Assemble into matrix a i ;

[0079] (e) Sample matrix construction: The displacement field matrix a under different load conditions is constructed... i Assemble into sample matrix A.

[0080] S103. Construct a reduced-order basis matrix based on the sample matrix, and project the sample matrix using the reduced-order basis matrix to construct a reduced-order model.

[0081] Constructing a reduced-order basis matrix based on the sample matrix specifically includes:

[0082] Construct a correlation matrix based on the sample matrix;

[0083] Construct the Lagrange function based on the correlation matrix and find its extrema to obtain the eigenvalues ​​and eigenvectors of the correlation matrix. Sort the eigenvalues ​​in descending order and select the first N eigenvectors to construct a reduced-order basis matrix.

[0084] For example, the method of separation of variables is used to decompose the sample matrix A into two parts that are independent of time and space, namely a series of coefficients α that are only related to time. i (t) and a series of spatially orthogonal basis functions β i The product of (x). Then the sample matrix A can be expressed as: Construct its correlation matrix B based on sample matrix A: B = A T A.

[0085] By using the Lagrange method, the Lagrange multiplier δ is introduced. i Construct the Lagrange function and find its extrema. According to the matrix differentiation rule, we can obtain: β i T Bβ i =δ i Among them, the Lagrange multiplier δ i Let β be the eigenvalue of the real symmetric matrix B. i Let be the eigenvectors of a real symmetric matrix B. Arrange the eigenvalues ​​in descending order, and select the eigenvectors corresponding to the N largest eigenvalues ​​to form a reduced-order basis matrix C:

[0086] C = [c1, c2, ..., c N ],

[0087] The value of N is determined according to the following energy criterion:

[0088] N = argmin{I(N) ≥ γ}

[0089] In the formula, the γ value represents the energy proportion of the intrinsic orthogonal mode set retained after mode truncation in the original mode set. In this embodiment, γ≥99%.

[0090] S104. Reduce the order of the reduced model. When the computational accuracy of the reduced model meets the preset accuracy requirements, the global reduced basis matrix and the reduced model that meets the preset accuracy requirements are obtained.

[0091] In this embodiment, the calculation accuracy is calculated as follows:

[0092]

[0093] Among them, R 2 N represents the computational accuracy of the reduced-order model. n This indicates the number of samples used to verify accuracy. This represents the i-th displacement field obtained based on the reduced-order model. This represents the i-th displacement field obtained based on the full-order model. N represents n The average value of each displacement field.

[0094] For example, (a) accuracy assessment: calculating R 2 The value is used to evaluate the accuracy of the reduced-order model;

[0095] (b) Model optimization: If R 2 If the value is lower than the predetermined threshold of 0.99, the uniform sequence sampling method based on Monte Carlo and spatial reduction techniques is used to determine the new excitation load, and steps 2 and 3 are repeated until R... 2 The value meets the requirements;

[0096] (c) Global reduced basis matrix: The final global reduced basis matrix C will be used for subsequent model correction and state prediction.

[0097] S105. Reconstruct the stiffness matrix and mass matrix using the global reduced-order basis matrix.

[0098] In this embodiment, the stiffness matrix K POD for:

[0099]

[0100] Among them, P s s represents the number of stiffness matrices that make up the overall stiffness matrix. i G represents the number of modes involved in the stiffness matrix during the order reduction process. j (x i ) is the design variable x of the stiffness matrix i The relevant parts, Let C represent the stiffness matrix projected from the reduced basis matrix, and k represent the reduced basis matrix. is To design variables x with stiffness matrix i Irrelevant parts;

[0101] Mass matrix M POD for:

[0102]

[0103] Where, q i h represents the number of modes involved in the quality matrix during the order reduction process. j (y i ) Design variables y with the quality matrix i The relevant parts, Let m represent the mass matrix after projection through the reduced basis matrix. is Represents the design variable y with the quality matrix. i Irrelevant parts.

[0104] For example, the mass matrix and stiffness matrix of the structure are reconstructed using the global reduced-order basis matrix C:

[0105] The overall stiffness matrix K of the structure can be regarded as Ps The superposition of stiffness matrices of each region.

[0106]

[0107] Where, k i Let x be the stiffness matrix of the i-th region, and let x be the stiffness matrix of that region. i The function is as follows:

[0108]

[0109] In the formula, g j (x i ) is the design variable x of the stiffness matrix i Related parts; k is To design variables x with stiffness matrix i Irrelevant parts.

[0110] Therefore, the overall stiffness matrix of the structure can be obtained:

[0111]

[0112] Furthermore, the stiffness matrix K of the reduced-order model can be determined. POD The mass matrix M of the reduced-order model POD .

[0113] S106. Calculate the numerical frequencies and modes of the jacket platform using the reconstructed stiffness matrix and mass matrix. Construct an objective function by combining the acceleration response data of the jacket platform collected by the structural monitoring sensor. Adjust the correlation coefficients of the design variables of the objective function through an iterative optimization algorithm until the objective function converges, and obtain the corrected finite element model.

[0114] In this embodiment, the objective function is:

[0115]

[0116] Among them, f j f is the j-th order numerical frequency obtained from the reduced-order model. Ej Let be the j-th order numerical frequency obtained from indoor model experiments, r be the number of numerical modes matching the experimental mode, μ be the weighting coefficient, and MAC be the numerical frequency. j Let be the degree of similarity between the j-th order numerical mode and the experimental mode. Let θ be the j-th order mode obtained from the reduced-order model. j Let be the j-th mode obtained from the full-order model.

[0117] For example, to achieve real-time correction of the finite element model, it is necessary to calculate the numerical frequencies and modes of the structure using a reduced-order model, construct a correction objective function, and iteratively optimize it until convergence. The specific implementation steps are as follows:

[0118] (a) Numerical frequency and modal calculations: using a reduced-order model K POD and M POD Calculate the numerical frequencies and modes of the structure;

[0119] (b) Test data acquisition: Acceleration response data was acquired using structural monitoring and sensing equipment based on the on-site jacket platform;

[0120] (c) Objective function construction: Define the objective function E to evaluate the effect of model correction.

[0121] (d) Iterative optimization: Adjust the coefficients g related to the design variables using gradient descent or other optimization algorithms. j (x i ) and h j (y i Gradually reduce the value of the objective function E until convergence.

[0122] In this embodiment, based on the initial structural parameters of the on-site jacket platform, a finite element model is constructed using ANSYS finite element analysis software; the main modal characteristics are preserved using POD technology, and the mass matrix and stiffness matrix of the structure are reconstructed through reduced-order basis matrices; the objective function for correcting the finite element model is calculated, and the model is corrected in real time through iterative iteration until convergence.

[0123] S107. Obtain the reduced-order finite element model based on the corrected finite element model, and apply the time series data to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time.

[0124] In this embodiment, a reduced-order finite element model is obtained based on the modified finite element model, and time series data is applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time. Specifically, this includes:

[0125] The marine environmental load is obtained by inverting the strain response data of the on-site jacket platform structure; the marine environmental load is input into a pre-trained LSTM model, and the pre-trained LSTM model outputs time series data containing future loads.

[0126] The modified finite element model is reduced in order using a method that meets the preset accuracy requirements, resulting in a reduced-order finite element model. Time series data is then applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time.

[0127] For example, marine environmental loads are derived from strain response data of an on-site platform structure. Strain sensors are installed at key locations on the jacket platform to collect strain response data in real time. The data acquisition frequency should be high enough to capture the platform's dynamic response. The acquired data is preprocessed, including filtering and denoising, to ensure accuracy and reliability. The load coefficient matrix is ​​solved using the least squares method from the strain mode matrix. This load coefficient matrix reflects the response characteristics of the local model under external loads. Then, based on the obtained load coefficient matrix, the marine environmental loads on the overall model are derived.

[0128] For example, preprocessing of marine environmental loads, including data cleaning, standardization, and normalization, is performed to improve the accuracy and stability of the prediction model. An LSTM neural network model is used to construct a 24-hour short-term dynamic prediction model for marine environmental loads. Historical load data and on-site monitoring data are used for model training to ensure good prediction accuracy. The trained model is then used to predict marine environmental loads for the next 24 hours, generating time-series data for future loads.

[0129] In this embodiment, the future operating state trend includes time series diagrams and predicted values ​​of displacement and strain; the method further includes:

[0130] The future operating status trend is displayed through a graphical interface. When the future operating status trend exceeds the safety threshold, an alarm mode is activated and an alarm message is generated. The alarm message includes the future fault occurrence time, specific location, cause of the fault, and handling measures.

[0131] The future operational trend exceeds the safety threshold, specifically including:

[0132] When the first strut of the jacket platform yields, the overall displacement value is set as the first-level warning value. The displacement value of the first-level warning value is determined based on the overall displacement when the strut yields. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements.

[0133] When the first leg of the jacket platform yields, the overall displacement value is set as the secondary warning value. The displacement value of the secondary warning value is determined based on the overall displacement when the leg yields. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements.

[0134] When a node of the jacket platform fails, the overall displacement value is set as the level three warning value. The displacement value of the level three warning value is determined based on the overall displacement when the node fails. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements.

[0135] The alarm levels in the alarm mode include Level 1, Level 2, and Level 3.

[0136] For example, environmental loads are applied to the real-time corrected finite element model of the jacket platform, and model order reduction techniques are used to predict the platform's future operating status trend in the short term, specifically including the platform's maximum displacement and maximum strain within the next 24 hours. A time series diagram of the prediction results is plotted to visually illustrate the platform's status trend. The specific implementation steps are as follows:

[0137] (a) Load application: The predicted environmental loads are applied to the reduced-order finite element model;

[0138] (b) State prediction: using the modified reduced-order model K POD and M POD The displacement and strain fields of the platform over the next 24 hours are solved using the dynamic equations. The dynamic equations can be expressed as:

[0139]

[0140] In the formula, C POD The damping matrix of the reduced-order model can be determined through experimental or empirical data;

[0141] (c) Prediction results display: The prediction results, including time series diagrams and predicted values ​​of displacement and strain, are displayed in real time through a graphical interface;

[0142] (d) Security Threshold Assessment: An alarm is triggered when the prediction result exceeds a preset security threshold. Specific thresholds are as follows:

[0143] When the first strut of the jacket structure yields, the overall displacement value is set as the first-level warning value. The displacement value for the first-level warning should be determined based on the overall displacement when the strut yields, and usually includes the maximum absolute value of the displacement in the x and y directions. When the first leg of the jacket structure yields, the overall displacement value is set as the second-level warning value. The displacement value for the second-level warning should be determined based on the overall displacement when the leg yields, and usually includes the maximum absolute value of the displacement in the x and y directions. When a node of the jacket structure fails, the overall displacement value is set as the third-level warning value. The displacement value for the third-level warning should be determined based on the overall displacement when the node fails, and usually includes the maximum absolute value of the displacement in the x and y directions.

[0144] In this embodiment, a real-time corrected finite element model is used to predict the platform's state, considering the platform's dynamic characteristics and environmental influencing factors. The prediction results, including time-series diagrams and predicted values ​​of displacement and strain, are displayed in real time through a graphical interface. When the prediction result exceeds a preset safety threshold, the system automatically generates an alarm. The alarm levels are divided into three levels: Level 1 (yellow), Level 2 (orange), and Level 3 (red). Users can set different alarm levels and thresholds according to actual conditions. The system automatically generates detailed alarm information, including the future fault occurrence time, specific location, and the exceeded value. The alarm information format is as follows:

[0145] Future failure time: accurate to the second, e.g., 2025-03-22 18:30:44;

[0146] Specific location: The exact location where the failure occurred on the on-site jacket platform, such as support structure connection node A;

[0147] Exceeding value: the specific value and the degree to which it exceeds the threshold;

[0148] The alert message is displayed through the user interface and sent to relevant personnel via SMS or email to remind them to take preventative measures. The message includes:

[0149] Alarm Summary: Briefly describe the main information of the alarm, such as "The displacement of connection node A of the platform support structure exceeds 15% of the design value";

[0150] Detailed links: Links to the user interface to allow relevant personnel to quickly view alarm information.

[0151] based on Figure 1 The method for real-time correction and status prediction of a jacket platform based on a reduced-order model, as shown in this invention, constructs a highly accurate digital twin synchronized with the physical jacket platform by combining model reduction technology, real-time model correction methods, and machine learning algorithms. This digital twin, based on real-time monitoring data and historical data, can assess the platform's future health status and predict potential safety risks. This significantly improves the operational safety and reliability of the jacket platform, provides a scientific basis for platform health management in complex marine environments, and has significant application value and innovation.

[0152] When applying the real-time correction and state prediction method for the jacket platform based on the reduced-order model provided in this manual, it is not necessary to consider... Figure 1 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this manual does not impose any restrictions on it.

[0153] The above describes one or more embodiments of a method for real-time correction and state prediction of a jacket platform based on a reduced-order model, as provided in this specification. Based on the same approach, this specification also provides a corresponding system for real-time correction and state prediction of a jacket platform based on a reduced-order model, such as... Figure 2 As shown.

[0154] Figure 2 This specification provides a schematic diagram of a real-time correction and state prediction system for a jacket platform based on a reduced-order model, including:

[0155] The finite element model building module is specifically used to build a finite element model of the jacket platform based on the initial structural parameters of the jacket platform.

[0156] The reduced-order model construction module is specifically used to apply at least two load conditions sequentially to the finite element model of the jacket platform, extract the displacement field at different times under each load condition, combine the displacement field into a sample matrix, construct a reduced-order basis matrix based on the sample matrix, project the sample matrix using the reduced-order basis matrix to construct a reduced-order model, reduce the order of the reduced-order model, and when the calculation accuracy of the reduced-order model meets the preset accuracy requirements, obtain the global reduced-order basis matrix and the reduced-order model that meets the preset accuracy requirements.

[0157] The model correction module is specifically used to reconstruct the stiffness matrix and mass matrix using the global reduced-order basis matrix; the numerical frequency and mode of the jacket platform are calculated using the reconstructed stiffness matrix and mass matrix; the objective function is constructed by combining the acceleration response data of the jacket platform collected by the structural monitoring sensor equipment; the correlation coefficient of the design variables of the objective function is adjusted by the iterative optimization algorithm until the objective function converges, and the corrected finite element model is obtained.

[0158] The state prediction module is specifically used to obtain a reduced-order finite element model based on the corrected finite element model, and apply time series data to the reduced-order finite element model to predict the future operating state trend of the jacket platform in real time.

[0159] Specific limitations regarding the real-time correction and state prediction system for jacket platforms based on reduced-order models can be found in the limitations of the real-time correction and state prediction method for jacket platforms based on reduced-order models mentioned above, and will not be repeated here. Each module in the aforementioned real-time correction and state prediction system for jacket platforms based on reduced-order models can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0160] This specification also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 A method for real-time correction and state prediction of the jacket platform based on a reduced-order model is provided.

[0161] This instruction manual also provides Figure 3 The schematic diagram of the computer device shown is as follows: Figure 3 At the hardware level, the computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to achieve the above-mentioned functions. Figure 1 A method for real-time correction and state prediction of the jacket platform based on a reduced-order model is provided.

[0162] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0163] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for real-time correction and state prediction of a jacket platform based on a reduced-order model, characterized in that, include: Based on the initial structural parameters of the jacket platform, a finite element model of the jacket platform is established; Based on the load variation characteristics of the historical load of the jacket platform in different time periods, the historical load is divided into at least two load conditions; Constructing a reduced-order basis matrix based on the sample matrix specifically includes: Construct a correlation matrix based on the sample matrix; Construct the Lagrange function based on the correlation matrix and find its extreme value to obtain the eigenvalues ​​and eigenvectors of the correlation matrix. Sort the eigenvalues ​​in descending order and select the first N eigenvectors to construct a reduced-order basis matrix. Based on the revised finite element model, a reduced-order finite element model is obtained, and time series data is applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time. Specifically, this includes: The marine environmental load is obtained by inverting the strain response data of the on-site jacket platform structure; the marine environmental load is input into a pre-trained LSTM model, and the pre-trained LSTM model outputs time series data containing future loads. The modified finite element model is reduced in order using a method that meets the preset accuracy requirements, resulting in a reduced-order finite element model. Time series data is then applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time. At least two load conditions are applied sequentially to the finite element model of the jacket platform, and the displacement field at different times under each load condition is extracted and combined into a sample matrix. A reduced-order basis matrix is ​​constructed based on the sample matrix, and the sample matrix is ​​projected onto the reduced-order basis matrix to construct a reduced-order model; The order of the reduced model is reduced. When the computational accuracy of the reduced model meets the preset accuracy requirements, the global reduced basis matrix and the reduced model that meets the preset accuracy requirements are obtained. Reconstruct the stiffness and mass matrices using a global reduced-order basis matrix; The numerical frequencies and modes of the jacket platform are calculated using the reconstructed stiffness matrix and mass matrix. The objective function is constructed by combining the acceleration response data of the jacket platform collected by the structural monitoring sensor. The correlation coefficients of the design variables of the objective function are adjusted by iterative optimization algorithm until the objective function converges, and the corrected finite element model is obtained. Based on the corrected finite element model, a reduced-order finite element model is obtained, and time series data is applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time.

2. The method for real-time correction and state prediction of a jacket platform based on a reduced-order model as described in claim 1, characterized in that, Sample matrix A = [a1, a2, ..., a i ], where a i For different times t j displacement field u(x) i ,t j The matrix assembled from these components. x i (i = 1, 2, ..., M) represents the i-th grid node, t j (j=1,2,…,N) represents the j-th time node of the finite element model of the jacket platform; The calculation method for accuracy is as follows: Among them, R 2 N represents the computational accuracy of the reduced-order model. n This indicates the number of samples used to verify accuracy. This represents the i-th displacement field obtained based on the reduced-order model. This represents the i-th displacement field obtained based on the full-order model. N represents n The average value of each displacement field.

3. The method for real-time correction and state prediction of a jacket platform based on a reduced-order model as described in claim 1, characterized in that, Stiffness matrix K POD for: Among them, P s s represents the number of stiffness matrices that make up the overall stiffness matrix. i G represents the number of modes involved in the stiffness matrix during the order reduction process. j (x i ) is the design variable x of the stiffness matrix i The relevant parts, Let C represent the stiffness matrix projected from the reduced basis matrix, and k represent the reduced basis matrix. is To design variables x with stiffness matrix i Irrelevant parts; Mass matrix M POD for: Where, q i h represents the number of modes involved in the quality matrix during the order reduction process. j (y i ) Design variables y with the quality matrix i The relevant parts, Let m represent the mass matrix after projection through the reduced basis matrix. is Represents the design variable y with the quality matrix. i Irrelevant parts.

4. The method for real-time correction and state prediction of a jacket platform based on a reduced-order model as described in claim 1, characterized in that, The objective function E is: Among them, f j f is the j-th order numerical frequency obtained from the reduced-order model. Ej Let be the j-th order numerical frequency obtained from indoor model experiments, r be the number of numerical modes matching the experimental mode, μ be the weighting coefficient, and MAC be the numerical frequency. j Let be the degree of similarity between the j-th order numerical mode and the experimental mode. Let j be the j-th order mode obtained from the reduced-order model. Let be the j-th mode obtained from the full-order model.

5. The method for real-time correction and state prediction of a jacket platform based on a reduced-order model as described in claim 1, characterized in that, Future operating status trends include time series diagrams and predicted values ​​for displacement and strain; The method also includes: The system displays future operational trends through a graphical interface. When these trends exceed safety thresholds, an alarm mode is activated, generating alarm information. This alarm information includes the time of the future fault, its specific location, the cause of the fault, and the corresponding handling measures.

6. The method for real-time correction and state prediction of a jacket platform based on a reduced-order model as described in claim 5, characterized in that, The future operational trend exceeds the safety threshold, specifically including: When the first strut of the jacket platform yields, the overall displacement value is set as the first-level warning value. The displacement value of the first-level warning value is determined based on the overall displacement when the strut yields. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements. When the first leg of the jacket platform yields, the overall displacement value is set as the secondary warning value. The displacement value of the secondary warning value is determined based on the overall displacement when the leg yields. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements. When a node of the jacket platform fails, the overall displacement value is set as the level three warning value. The displacement value of the level three warning value is determined based on the overall displacement when the node fails. The overall displacement includes the maximum absolute value of the horizontal and vertical displacements. The alarm levels in the alarm mode include Level 1, Level 2, and Level 3.

7. A real-time correction and state prediction system for a jacket platform based on a reduced-order model, characterized in that, include: The finite element model building module is specifically used to build a finite element model of the jacket platform based on the initial structural parameters of the jacket platform. The load division module is specifically used to divide the historical load of the jacket platform into at least two load conditions based on the load change characteristics of the historical load in different time periods. The reduced basis matrix construction module is specifically used to construct reduced basis matrices based on sample matrices, and includes: A correlation matrix is ​​constructed based on the sample matrix; a Lagrangian function is constructed based on the correlation matrix and its extreme values ​​are found to obtain the eigenvalues ​​and eigenvectors of the correlation matrix. The eigenvalues ​​are arranged in descending order, and the first N eigenvectors are selected to construct a reduced-order basis matrix; a reduced-order finite element model is obtained based on the corrected finite element model, and time series data is applied to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time. Specifically, this includes: inverting the marine environmental load based on the strain response data of the jacket platform structure to obtain the marine environmental load; inputting the marine environmental load into a pre-trained LSTM model, and outputting time series data containing the future load through the pre-trained LSTM model; reducing the order of the corrected finite element model using a reduction method that meets the preset accuracy requirements to obtain a reduced-order finite element model; and applying time series data to the reduced-order finite element model to predict the future operating status trend of the jacket platform in real time. The reduced-order model construction module is specifically used to apply at least two load conditions sequentially to the finite element model of the jacket platform, extract the displacement field at different times under each load condition, combine the displacement field into a sample matrix, construct a reduced-order basis matrix based on the sample matrix, project the sample matrix using the reduced-order basis matrix to construct a reduced-order model, reduce the order of the reduced-order model, and when the calculation accuracy of the reduced-order model meets the preset accuracy requirements, obtain the global reduced-order basis matrix and the reduced-order model that meets the preset accuracy requirements. The model correction module is specifically used to reconstruct the stiffness matrix and mass matrix using the global reduced-order basis matrix; the numerical frequency and mode of the jacket platform are calculated using the reconstructed stiffness matrix and mass matrix; the objective function is constructed by combining the acceleration response data of the jacket platform collected by the structural monitoring sensor equipment; the correlation coefficient of the design variables of the objective function is adjusted by the iterative optimization algorithm until the objective function converges, and the corrected finite element model is obtained. The state prediction module is specifically used to obtain a reduced-order finite element model based on the corrected finite element model, and apply time series data to the reduced-order finite element model to predict the future operating state trend of the jacket platform in real time.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the method described in any one of claims 1 to 6.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in any one of claims 1 to 6.

Images