Wave load damage assessment method for semi-floating floating platform structures

Abstract

This invention provides a method for assessing wave load damage to a semi-floating structure of a floating platform, relating to the field of fatigue analysis technology for offshore floating platform structures. The method includes discretizing the wave spectrum into multiple frequency points using an energy segmentation method, inputting these points into a trained wave load model to obtain wave load time history data; calculating the Reynolds number and structural compliance coefficient, and further calculating the dynamic coupling coefficient accordingly; performing a Discrete Fourier Transform (DFT) on the wave load time history data, interpolating the initial frequency response parameters, and performing an Inverse DFT on the stress frequency domain response parameters to obtain the corresponding time-domain stress time history parameters. By introducing a wave load model, the discretized wave spectrum is directly mapped to wave load time history data, achieving efficient conversion from the load frequency domain to the stress time domain through DFT and Inverse DFT.

Description

Technical Field

[0001] This invention relates to the field of fatigue analysis technology for offshore floating platform structures, specifically a method for assessing wave load damage to semi-floating structures of floating platforms. Background Technology

[0002] Floating platforms, such as semi-submersible platforms, tension leg platforms, and SPAR platforms, are crucial infrastructure for deep-sea oil and gas development and offshore wind power, requiring continuous service at sea for 20 to 30 years. During long-term service, the platform structure is continuously subjected to alternating environmental loads such as waves, wind, and currents, making fatigue damage one of its main failure modes. Currently, fatigue analysis of floating platform structures mainly employs two categories of methods: time-domain methods and frequency-domain methods. Time-domain methods, such as rainflow counting combined with finite element time-history analysis, offer high accuracy but involve enormous computational demands, making it difficult to meet the needs of rapid engineering assessment and real-time monitoring. Frequency-domain methods, based on power spectral density for fatigue damage prediction, offer high computational efficiency, but existing frequency-domain methods typically assume the structure is in a fixed, fully floating equilibrium state, failing to consider the effects of time-varying wet surfaces, dynamic changes in boundary conditions, and nonlinear effects such as wave impact under semi-floating conditions. In the prior art, document CN116796591A proposes a method to establish a three-dimensional static and dynamic analysis model based on Abaqus finite element software, including the wind turbine, tower, monopile foundation, surrounding soil, and external environment. The model incorporates a Mohr-Coulomb model to simulate nonlinear pile-soil contact, applies aerodynamic and Rayleigh damping, calculates blade wind loads based on blade element momentum theory, and generates a wave model using Jonswap spectrum combined with MacCamy-Fuchs diffraction correction. Finally, wave forces are solved using the Morison equation, achieving simulation of the dynamic response of a fixed offshore wind power structure under wind, wave, and current conditions. However, this method is only applicable to fixed monopile foundations and cannot handle strongly nonlinear fluid-structure interaction problems such as time-varying wet surfaces, dynamic changes in boundary conditions, and wave slamming on floating platforms in a semi-floating state. First, the method's analysis scenarios are limited to service conditions where the foundation is completely fixed. Second, it relies on empirical formulas such as the Morison equation to repeatedly solve for wave forces, resulting in limited accuracy and high computational cost for large-scale floating structures under complex sea conditions, and failing to achieve efficient and intelligent mapping from wave spectrum to load time history. Third, it does not establish a joint analysis framework from frequency domain load to stress response, but instead directly solves for dynamic response in the time domain, limiting analysis efficiency. Finally, the fatigue assessment part of this method lacks consideration for the time-varying degradation of material properties caused by corrosive environments, failing to reflect the continuous decay of fatigue strength throughout the structure's lifespan, and also lacks a cumulative damage early warning mechanism, making it difficult to meet the needs of structural fatigue life prediction and safety assessment for floating platforms in complex marine environments. Therefore, a wave load damage assessment method for semi-floating floating platform structures is urgently needed.

[0003] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] The purpose of this invention is to provide a method for assessing wave load damage to a semi-floating floating platform structure, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for assessing wave load damage to semi-floating structures of floating platforms, comprising the following specific steps: S1: Obtain wave environment parameters of the target sea area, extract geometric and material parameters of key parts of the floating platform, establish a finite element model of the floating platform, and perform harmonic response analysis on the finite element model to obtain preliminary frequency response parameters. S2: Based on the wave environment parameters, construct a JONSWAP spectrum model and calculate the wave spectrum accordingly. Use the energy segmentation method to discretize the wave spectrum into multiple frequency points, construct a wave load model based on LSTM, and input the discretized wave spectrum into the trained wave load model to obtain wave load time history data. Based on the wave environment parameters, geometric parameters, and material parameters, calculate the Reynolds number and structural flexibility coefficient, and further calculate the dynamic coupling coefficient accordingly. S3: Perform Discrete Fourier Transform on the wave load time history data, interpolate the initial frequency response parameters to make them consistent with the frequency points of the wave load time history data after Discrete Fourier Transform, calculate the stress frequency domain response parameters based on the dynamic coupling coefficient and the interpolated frequency response parameters, and then perform Inverse Discrete Fourier Transform on the stress frequency domain response parameters to obtain the corresponding time domain stress time history parameters. S4: Construct a time-varying corrosion factor based on the material parameters, use the rainflow counting method to extract the stress amplitude and stress cycle number from the time-domain stress time history parameters, and use the time-varying corrosion factor as a correction term to construct the corrected stress cycle number; calculate the cumulative damage based on the stress cycle number, and issue an alarm when the cumulative damage is greater than or equal to the preset damage value.

[0006] Furthermore, the geometric and material parameters of the floating platform are extracted, specifically: Wave environment parameters, including significant wave height and spectral peak period, are obtained from a marine meteorological database. At the same time, geometric and material parameters of key parts of the floating platform are also obtained. The key parts are the connection between the column and the pontoon. The geometric parameters include the column diameter and plate thickness, and the material parameters include the elastic modulus, Poisson's ratio, and corrosion rate. A finite element model of the floating platform is established, and harmonic response analysis is performed on the finite element model. Specifically, a unit harmonic excitation along the wave propagation direction is applied to the floating platform, the stress response amplitude of the key parts at each frequency point of the unit harmonic excitation is extracted, and the ratio of the stress response amplitude to the excitation amplitude is calculated to obtain preliminary frequency response parameters.

[0007] Furthermore, a JONSWAP spectral model is constructed and the wave spectrum is calculated accordingly, specifically as follows: Based on wave environment parameters, construct the JONSWAP spectral model: in, Indicates the peak factor; Indicates the significant wave height; Indicates the period of the spectral peak; Indicates angular frequency; Indicates angular frequency as Wave spectrum at time; This represents the peak shape parameter.

[0008] Furthermore, the wave spectrum is discretized into multiple frequency points using the energy segmentation method, specifically: The total wave energy is calculated as follows: in, This represents the total wave energy; Indicates the angular frequency of the spectral peak; Indicates the discrete lower limit frequency; Indicates the discrete upper limit frequency; Divide the total wave energy into equal parts to obtain the unit energy value: in, Indicates the value per unit of energy; Indicates the number of energy portions; Solve for the boundary values ​​at each frequency point: in, Indicates the energy boundary frequency; Indicates the index of the number of energy segments; The final discrete frequency point set is: ,in, It represents a set of discrete frequencies.

[0009] Furthermore, the specific logic for acquiring wave load time history data is as follows: Multiple sets of discrete wave spectra and corresponding discrete wave load time sequence data are obtained from the HELOFOW project database. The discrete wave spectra are used as the input of the wave load model, and the discrete wave load time sequence data are used as the output of the wave load model. The wave load model is trained. When the error between the predicted output result of consecutive preset training rounds and the true value of the discrete wave load time sequence data does not exceed the set prediction error, the training is considered complete. The wave spectrum, discretized into multiple frequency points, is input into the trained wave load model to obtain the corresponding discrete wave load time program. The wave load model is based on an LSTM model architecture, specifically including an input layer, a first LSTM layer, a second LSTM layer, a third LSTM layer, a first dropout layer, a second dropout layer, a third dropout layer, a fully connected layer, and an output layer. The input data for the input layer is a wave spectrum discretized into multiple frequency points, and the output data of the input layer is input to the first LSTM layer. The first LSTM layer is a first Long Short-Term Memory (LSTM) network layer containing 128 memory units, and the output data of the first LSTM layer is input to both the second LSTM layer and the first dropout layer. The second LSTM layer is a second LSTM layer containing 128 memory units. The third LSTM layer is a Long Short-Term Memory (LSTM) network layer. The output data of the second LSTM layer and the output data of the first discard layer are input into the third LSTM layer. The third LSTM layer is a third LSTM network layer containing 128 memory units. The output data of the third LSTM layer is input into the second discard layer. The output data of the second discard layer is input into the fully connected layer. The output data of the fully connected layer is input into the third discard layer. The output data of the third discard layer is input into the output layer. The output data of the output layer is a discrete wave load time sequence. The discard rate of the first, second, and third discard layers is 0.2.

[0010] Furthermore, the dynamic coupling coefficient is calculated, specifically as follows: Calculate the Reynolds number and structural compliance coefficient based on wave environment parameters, geometric parameters, and material parameters: in, Indicates the diameter of the column; Indicates the density of seawater; Indicates the significant wave height; Indicates the period of the spectral peak; Represents the Reynolds number; Indicates the elastic modulus of the material in critical components; Indicates the plate thickness of key components; Indicates the Poisson's ratio of the material in the critical component; Indicates the structural flexibility coefficient; Indicates the dynamic viscosity of seawater; Represents gravitational acceleration; Calculate the dynamic coupling coefficient based on the Reynolds number and structural flexibility coefficient: in, This represents the dynamic coupling coefficient.

[0011] Further, the interpolated frequency response parameters are calculated as follows: Perform Discrete Fourier Transform on wave load time history data: in, In the program sequence representing the discrete wave load output by the wave load model, the first... Load values ​​at each time point; This indicates the frequency at which the discrete Fourier transform is applied. Wave load frequency domain data at the location; Indicates the first A discrete angular frequency point; Indicates a discrete time point index; Indicates the total length of the program sequence when dealing with wave loads; Indicates the first The time values ​​corresponding to each discrete time point; Indicates the index of discrete angular frequency points; The initial frequency response parameters are interpolated to match... By keeping the frequency points consistent, the interpolated frequency response parameters are obtained: for For each target frequency point, first determine whether the target frequency point is within the known frequency range of the preliminary frequency response parameters. If the target frequency point is lower than the known lowest frequency, then directly take the parameter value corresponding to the known lowest frequency point as the interpolation result of the target frequency point; if the target frequency point is higher than the known highest frequency, then directly take the parameter value corresponding to the known highest frequency point as the interpolation result of the target frequency point. If the target frequency point is within the known frequency range, then search from left to right in the sequence of known frequency points to find the first known frequency point that is greater than or equal to the target frequency point, and simultaneously take the previous known frequency point. These two points are the two known frequency points adjacent to the target frequency point. Then calculate the distance from the target frequency point to the previous known frequency point, and the total distance between the two known frequency points. The interpolated parameter value of the target frequency point can be obtained by adding the parameter value of the previous known frequency point to the change in the parameter value over the total distance and multiplying it by the distance ratio. The distance ratio is equal to the distance from the target frequency point to the previous known frequency point divided by the distance between the two known frequency points.

[0012] Furthermore, the time-domain stress time history parameters are calculated as follows: Based on the dynamic coupling coefficient and the interpolated frequency response parameters, the stress frequency domain response parameters are calculated: in, Indicates at angular frequency Stress frequency domain response parameters at the location; This represents the frequency response parameter after interpolation; Performing an inverse discrete Fourier transform on the stress frequency domain response yields the corresponding time-domain stress time-history parameters: in, After the inverse discrete Fourier transform, at the... Time-domain stress time history parameters at discrete time points.

[0013] Further, the cumulative damage is calculated as follows: Based on material parameters, a time-varying corrosion factor is constructed: in, express The time-varying corrosion factor at any given moment; Indicates the corrosion rate of the floating platform material; Represents a time variable; The stress amplitude and stress cycle number were extracted from the time-domain stress time history parameters using the rainflow counting method, and the time-varying corrosion factor was used as a correction term to construct the corrected SN curve. in, Indicates the first The amplitude of the second stress cycle; Indicates the first Secondary stress cycle amplitude The number of stress cycles that a floating platform can withstand under the action of force; The slope of the SN curve represents the floating platform. This represents the intercept constant of the SN curve for a floating platform; Calculate cumulative damage based on the number of stress cycles: in, Indicates cumulative damage; This represents the total number of stress cycle sets extracted by the rainflow counting method; Indicates the stress cycle number index; When the accumulated damage is greater than or equal to the preset damage value, it indicates that the floating platform is in a safe state; otherwise, it indicates that the floating platform is in a dangerous state and an alarm will be issued immediately.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This scheme establishes an end-to-end mapping from discrete wave spectrum to wave load time history using an LSTM neural network, replacing the traditional iterative solution based on the Morison equation. This significantly improves load calculation efficiency while maintaining accuracy. Simultaneously, the energy segmentation method ensures the representativeness of discrete frequency points for wave energy distribution. By introducing a dynamic coupling coefficient based on Reynolds number and structural compliance coefficient, the nonlinear interaction between fluid and structure in a semi-floating state is quantitatively characterized, overcoming the shortcomings of traditional linear frequency domain methods in terms of insufficient accuracy under wet surface time-varying and slamming load conditions, making stress response calculation more closely reflect actual working conditions. An efficient conversion framework from load frequency domain to stress time domain is constructed through discrete Fourier transform and inverse Fourier transform, retaining the high computational efficiency of frequency domain methods while providing the necessary time-domain stress data for rainflow counting, balancing computational efficiency and damage assessment accuracy. A time-varying corrosion factor is introduced to correct the SN curve, considering the continuous decay of material corrosion on fatigue performance, avoiding the problem of overly optimistic fatigue life in the later stages of corrosion caused by a fixed SN curve, improving the reliability of cumulative damage assessment throughout the entire life cycle, and achieving real-time monitoring of structural safety through a threshold early warning mechanism. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a graph showing the relationship between the Reynolds number and the corresponding dynamic coupling coefficient. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0017] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example: Please see Figures 1-2 The present invention provides a technical solution: A method for assessing wave load damage to semi-floating structures of floating platforms, comprising the following specific steps: S1: Obtain wave environment parameters of the target sea area, extract geometric and material parameters of key parts of the floating platform, establish a finite element model of the floating platform, and perform harmonic response analysis on the finite element model to obtain preliminary frequency response parameters. In this embodiment, wave spectrum data of the target sea area is acquired, and the geometric and material parameters of the floating platform are extracted, specifically as follows: Wave environment parameters, including significant wave height and spectral peak period, are obtained from a marine meteorological database. At the same time, geometric and material parameters of key parts of the floating platform are also obtained. The key parts are the connection between the column and the pontoon. The geometric parameters include the column diameter and plate thickness, and the material parameters include the elastic modulus, Poisson's ratio, and corrosion rate. A finite element model of the floating platform is established, and harmonic response analysis is performed on the finite element model. Specifically, a unit harmonic excitation along the wave propagation direction is applied to the floating platform, the stress response amplitude of the key parts at each frequency point of the unit harmonic excitation is extracted, and the ratio of the stress response amplitude to the excitation amplitude is calculated to obtain preliminary frequency response parameters.

[0019] In the above process, the connection between the column and the pontoon is the area with the most drastic change in cross-sectional stiffness in the overall structure of the floating platform. At the same time, this location is simultaneously subjected to the axial force and bending moment from the column, as well as the buoyancy and wave slamming load from the pontoon, resulting in significant stress concentration under wave cyclic load. Using this location as a key monitoring point can capture the impact of load redistribution caused by wet surface changes in the semi-floating state on fatigue hotspots at the connection, and can also use the stress response amplitude at this location as a characterization of frequency response parameters. This is because the local stress at this location is most sensitive to global wave excitation, and its response characteristics can represent the dynamic transmission characteristics of the entire platform under unit harmonic excitation, thus providing the most conservative and representative fatigue hotspot data for subsequent stress frequency domain calculations and cumulative damage assessment. By directly obtaining the effective wave height and spectral peak period from the marine meteorological database as wave environment input, and simultaneously extracting the column diameter, plate thickness, elastic modulus, Poisson's ratio, and corrosion rate at the connection between the column and the pontoon as structural characteristic parameters, the key control parameters of environmental load characteristics and structural resistance are clearly separated, providing a physical basis for subsequent wave spectrum construction and LSTM load prediction. Based on this, by applying a unit harmonic excitation along the wave propagation direction to the finite element model of the floating platform, the stress response amplitude at each frequency point at the connection between the column and the pontoon is extracted, and the ratio of the stress response amplitude to the excitation amplitude is calculated to obtain preliminary frequency response parameters. The technical effect achieved is that the unit harmonic excitation makes the frequency response parameters reflect only the inherent dynamic characteristics of the structure itself, such as stiffness, mass, and damping distribution, and is independent of the specific load amplitude. Therefore, this preliminary frequency response parameter can be used as the structural transfer function to be multiplied with any wave load frequency domain data in subsequent steps, thus eliminating the need to repeatedly perform finite element harmonic response analysis for each sea state, significantly reducing the number of finite element calculations, and ensuring the speed and consistency of stress response calculation under different sea state conditions. S2: Based on the wave environment parameters, construct a JONSWAP spectrum model and calculate the wave spectrum accordingly. Use the energy segmentation method to discretize the wave spectrum into multiple frequency points, construct a wave load model based on LSTM, and input the discretized wave spectrum into the trained wave load model to obtain wave load time history data. Based on the wave environment parameters, geometric parameters, and material parameters, calculate the Reynolds number and structural flexibility coefficient, and further calculate the dynamic coupling coefficient accordingly. In this embodiment, the JONSWAP spectral model is constructed and the wave spectrum is calculated accordingly, specifically as follows: Based on wave environment parameters, construct the JONSWAP spectral model: in, Indicates the peak factor; Indicates the significant wave height; Indicates the period of the spectral peak; Indicates angular frequency; Indicates angular frequency as Wave spectrum at time; This represents the peak shape parameter.

[0020] In the above process, the dependent variable Indicates angular frequency as The wave spectrum at a given time, in its physical sense, is the wave energy distribution density within a unit frequency interval. The larger the value, the higher the energy contained in the wave at that frequency, and the stronger the excitation effect on the floating platform structure. Effective wave height (independent variable) Spectral peak period and peak factor Influence The physical mechanism lies in: significant wave height It characterizes the average height of the significant waves in the wave train. The larger the wave, the higher its total energy. and Proportional, reflecting the linear relationship between the square of wave height and wave energy; spectral peak period This determines the location where wave energy is concentrated in the frequency domain. The larger the value, the lower the spectral peak shifts ( ). (reduce), at the same time The increase through the denominator and exponential terms The combined effect of these factors narrows and elevates the spectral peak, reflecting the physical law that energy concentrates in lower frequencies after the wind and waves have fully developed; spectral peak factor Controlling the sharpness of spectral peaks, A larger value indicates a higher concentration of energy near the spectral peak frequency, and a stronger instability of the wave field; spectral peak shape parameter Take 0.07 (when (when) and 0.09 (when) (Time) controls the attenuation steepness of the low-frequency and high-frequency sides of the spectral peak respectively. The smaller the peak, the sharper it is; conversely, the larger the peak, the flatter it is. The rationality of this formula lies in the following: High-frequency attenuation term and Together, they ensured rapid energy decay in the high-frequency band and limited energy in the low-frequency band, consistent with physical observations in actual sea conditions where wave energy is concentrated near the spectral peak and high-frequency components dissipate rapidly; while By applying multiplicative correction only to the frequency band near the spectral peak, without affecting the asymptotic decay characteristics of high and low frequencies, this model retains the theoretical foundation of the Pierson-Moskowitz spectrum while also allowing for adjustment of the spectral peak factor. and peak shape parameters Adapting to sea conditions at different growth stages provides accurate energy distribution input for the subsequent energy segmentation method to discretize the wave spectrum into representative frequency points; from the perspective of positive and negative correlation, and They are strictly positively correlated. As the wave increases, the total wave energy increases quadratically. With spectral peak factor They show a positive correlation near the spectral peak. Increasing the frequency makes the spectral peak energy more concentrated and the peak value higher, but in the low-frequency and high-frequency ranges far from the spectral peak, The corrective effect is close to It has almost no impact; With angular frequency The relationship exhibits a unimodal characteristic of first increasing and then decreasing. near It reached its peak at that time. When far from the spectral peak frequency The rapid decay to near zero is consistent with the physical characteristic that actual wave energy is only distributed within a limited frequency band.

[0021] The wave spectrum is discretized into multiple frequency points using the energy segmentation method, specifically as follows: The total wave energy is calculated as follows: in, This represents the total wave energy; Indicates the angular frequency of the spectral peak; Indicates the discrete lower limit frequency; Indicates the discrete upper limit frequency; Divide the total wave energy into equal parts to obtain the unit energy value: in, Indicates the value per unit of energy; Indicates the number of energy portions; Solve for the boundary values ​​at each frequency point: in, Indicates the energy boundary frequency; Indicates the index of the number of energy segments; The final discrete frequency point set is: ,in, It represents a set of discrete frequencies.

[0022] In the above process, by using the continuous wave spectrum In limited frequency band The total wave energy is obtained by inner integration. Then, the total wave energy is divided equally. A unit of energy value is obtained. Then, by numerical integration, the energy boundary frequencies are solved in reverse. (satisfy Ultimately, this compresses the continuous spectrum into a form containing... A set of discrete frequency points The technical effect achieved by this method is that the energy segmentation method automatically densifies the frequency points in regions with high wave energy (i.e., large spectral values) and automatically sparses them in regions with low wave energy (i.e., small spectral values). This maximizes the preservation of the wave spectrum's energy distribution characteristics with a limited number of discrete frequency points, avoiding the problems of information loss near spectral peaks or redundant sampling in non-peak regions that occur with equal-interval frequency discretization. Simultaneously, and The cutoff range ensures that more than 99% of the total energy in the JONSWAP spectrum is covered, making the energy loss caused by discretization negligible. This provides compact and physically representative discrete frequency features for the subsequent LSTM model input, thereby significantly reducing the input complexity and training difficulty of the LSTM model while ensuring the accuracy of wave load prediction.

[0023] The specific logic for obtaining the wave load time history data is as follows: Multiple sets of discrete wave spectra and corresponding discrete wave load time sequence data are obtained from the HELOFOW project database. The discrete wave spectra are used as the input of the wave load model, and the discrete wave load time sequence data are used as the output of the wave load model. The wave load model is trained. When the error between the predicted output result of consecutive preset training rounds and the true value of the discrete wave load time sequence data does not exceed the set prediction error, the training is considered complete. The wave spectrum, discretized into multiple frequency points, is input into the trained wave load model to obtain the corresponding discrete wave load time program. The wave load model is based on an LSTM model architecture, specifically including an input layer, a first LSTM layer, a second LSTM layer, a third LSTM layer, a first dropout layer, a second dropout layer, a third dropout layer, a fully connected layer, and an output layer. The input data for the input layer is a wave spectrum discretized into multiple frequency points, and the output data of the input layer is input to the first LSTM layer. The first LSTM layer is a first Long Short-Term Memory (LSTM) network layer containing 128 memory units, and the output data of the first LSTM layer is input to both the second LSTM layer and the first dropout layer. The second LSTM layer is a second LSTM layer containing 128 memory units. The third LSTM layer is a Long Short-Term Memory (LSTM) network layer. The output data of the second LSTM layer and the output data of the first discard layer are input into the third LSTM layer. The third LSTM layer is a third LSTM network layer containing 128 memory units. The output data of the third LSTM layer is input into the second discard layer. The output data of the second discard layer is input into the fully connected layer. The output data of the fully connected layer is input into the third discard layer. The output data of the third discard layer is input into the output layer. The output data of the output layer is a discrete wave load time sequence. The discard rate of the first, second, and third discard layers is 0.2.

[0024] In the above process, multiple sets of discrete wave spectra and corresponding discrete wave load time sequence data are obtained from the HELOFOW project database. The LSTM model is trained using discrete wave spectra as input and discrete wave load time sequence as output. The aim is to utilize the long-range dependency modeling capability of the LSTM network for sequence data to establish an end-to-end mapping relationship from the frequency domain wave energy distribution (wave spectrum) to the time domain load evolution process (load time sequence), thereby replacing the traditional calculation process of repeatedly solving wave forces based on physical formulas such as the Morison equation. The LSTM model uses a three-layer stacked LSTM architecture (each layer containing 128 memory units), three dropout layers (all with a dropout rate of 0.2), and a fully connected layer. This allows the first LSTM layer to initially extract the energy distribution characteristics between different frequency points of the wave spectrum, and the second LSTM layer to further extract higher-order temporal dependencies and connect them with the output of the first dropout layer. The concatenated output is input into the third LSTM layer. This enhances the model's expressive power while mitigating the gradient vanishing problem through interlayer residual connections. The three dropout layers randomly shut down neurons with a probability of 0.2 during training to prevent the model from overfitting the training data, thereby improving generalization ability. The technical effect of this process is that after training and convergence, the trained LSTM model can directly map the wave spectrum, which is discretized into multiple frequency points, as input to obtain the corresponding discrete wave load time sequence. This eliminates the need to perform hydrodynamic numerical simulations for each sea state, significantly improving the efficiency of wave load calculation. At the same time, the multi-layer stacked structure and residual connections of the LSTM ensure the model's fitting accuracy for complex nonlinear mapping relationships. This allows the intelligently predicted load time history to retain the integrity of the wave spectrum frequency domain information and possess the physical rationality of the time domain load evolution process, providing a reliable and efficient load input for subsequent frequency domain stress response calculation. The method for determining the preset training rounds is based on observing the changing trend of the validation set loss function during training: Multiple sets of discrete wave spectra and corresponding discrete wave load time sequence data from the HELOFOW project database are proportionally divided into training, validation, and test sets (typically 70%, 15%, and 15%). At the end of each training round, the loss function value between the model's predicted output and the true value on the validation set is calculated. When the validation set loss function value no longer decreases for several consecutive rounds, this consecutive number of rounds is set as the preset round. The technical effect is to avoid the model overfitting the training data and losing its generalization ability due to too many training rounds, while preventing premature termination of the model before it has fully learned the mapping relationship due to too few rounds, ensuring that the model stops training at the position where its performance on the validation set is optimal. The method for determining the prediction error is based on the allowable deviation range for wave load time history prediction in engineering applications, such as peak load prediction error not exceeding ±5% or root mean square error not exceeding 10% of the standard deviation of the measured data. The reason for setting the dropout rate of the first dropout layer to 0.2 is that the first dropout layer is located between the first LSTM layer and the subsequent networks. Its role is to perform initial regularization on the wave spectrum frequency domain features extracted by the first LSTM layer before they enter the deeper networks. A dropout rate of 0.2 means that 20% of the neuron outputs are randomly set to zero during training. This proportion ensures that a sufficient proportion (80%) of the frequency domain features output by the first LSTM layer are retained and passed to the third LSTM layer, so as to ensure that the energy distribution features between different frequency points of the wave spectrum are not excessively destroyed. At the same time, the moderate random dropout forces the first LSTM layer to learn a more robust feature representation, preventing the model from becoming dependent on specific wave spectrum patterns in the training dataset.

[0025] Secondly, the reason for setting the dropout rate of the second dropout layer to 0.2 is that the second dropout layer is set between the output of the third LSTM layer and the fully connected layer. Its function is to perform intermediate layer regularization after the aggregation of temporal features in the deepest layer is completed and before entering the linear mapping. A dropout rate of 0.2 means that this layer randomly discards neurons in the high-dimensional temporal features output by the third LSTM layer with a 20% probability. This ratio ensures that the main components (80%) in the temporal features can be completely transmitted to the fully connected layer to ensure the accuracy of the load time history prediction. At the same time, by introducing random noise, complex co-adaptation relationships are prevented among the 128 memory units in the third LSTM layer, avoiding the model from relying too much on the output combination of specific memory units.

[0026] Finally, the reason for setting the dropout rate of the third dropout layer to 0.2 is that the third dropout layer is set between the output of the fully connected layer and the output layer. Its function is to perform pre-output regularization after the fully connected layer linearly maps the high-dimensional features extracted by LSTM to the target output dimension and before the program sequence when finally outputting discrete wave loads. A dropout rate of 0.2 means that 20% of the neurons output by the fully connected layer are randomly dropped. This proportion ensures that most of the effective information (80%) in the load features output by the fully connected layer is retained in the output layer to ensure prediction accuracy. At the same time, randomness is introduced to prevent the weight parameters of the fully connected layer from overfitting to the noise in the training data, and to ensure that the model can still stably output physically reasonable load time histories when faced with previously unseen discrete wave spectrum inputs.

[0027] The dynamic coupling coefficient is calculated as follows: Calculate the Reynolds number and structural compliance coefficient based on wave environment parameters, geometric parameters, and material parameters: in, Indicates the diameter of the column; Indicates the density of seawater; Indicates the significant wave height; Indicates the period of the spectral peak; Represents the Reynolds number; Indicates the elastic modulus of the material in critical components; Indicates the plate thickness of key components; Indicates the Poisson's ratio of the material in the critical component; Indicates the structural flexibility coefficient; Indicates the dynamic viscosity of seawater; Represents gravitational acceleration; In the above process, at the Reynolds number In the calculation formula, the dependent variable The ratio of inertial force to viscous force used to characterize fluid flow under wave action. A large value indicates that inertial forces dominate the flow, the fluid tends to be in a turbulent state, and pressure pulsation and vortex-induced vibration on the structural surface are enhanced. A smaller value indicates that viscous forces dominate, the flow tends to be laminar, and the viscous damping of the structure is relatively increased; in this formula, the effective wave height... and column diameter right They are positively correlated because An increase indicates an increase in the characteristic velocity of wave-like water particles. Increasing the characteristic scale representing the fluid flow increases both the inertial effect and the spectral peak period. right It is negatively correlated because An increase means a decrease in wave frequency and a decrease in the velocity of water particles, which relatively weakens the inertial effect and increases the density of seawater. and dynamic viscosity Influenced from the perspectives of inertial force and viscous force respectively , Increasing the constraint inertial force increases the inertial force. Increasing the constraint increases the viscous resistance; both can be controlled from positive and negative directions, respectively. The value of ; the rationality of the formula lies in the use of Approximate characteristic velocities of wave-like water particles The introduction of this makes the simple harmonic wave approximation With velocity amplitude ( For amplitude, (The magnitudes are consistent, and the column diameters are...) As a characteristic length, it conforms to the classical definition of flow around a cylinder, thus... The general form is appropriately expressed in the semi-floating wave load problem; In structural flexibility coefficient In the calculation formula, the dependent variable The ratio of the bending stiffness of the structure at the connection between the column and the pontoon of a floating platform to the characteristic quantity of hydrostatic pressure is used to characterize the structure. A larger value means that the structure is more flexible relative to fluid loads and is more likely to deform under wave action. A smaller value means a more rigid structure and a stronger ability to resist deformation; elastic modulus and plate thickness (by (appearance in form) It is negatively correlated because Increased resistance to deformation represents a greater ability of the material itself to resist deformation. Increasing the representative section's flexural stiffness increases cubically, both of which make the structure more rigid, while Poisson's ratio... right They are positively correlated because Increasing the column diameter makes the in-plane lateral contraction more significant, reduces the effective in-plane stiffness, and makes the structure exhibit greater compliance under the same load. (by (appearance in form) They are positively correlated because Increasing the characteristic scale representing the hydrostatic pressure of the fluid by a cube increases the fluid load on the structure, making it more compliant; the rationality of this formula lies in... Characterizing the bending stiffness (i.e., flexural stiffness) of thin plates, with The hydrostatic pressure characteristic of a floating platform in waves is characterized by the ratio of the hydrostatic pressure to the wave pressure. Dividing the two values ​​yields a dimensionless compliance coefficient, enabling floating platforms of different sizes and materials to have comparable compliance indices. This provides a basis for subsequently constructing dynamic coupling coefficients. It provides a unified physical base value.

[0028] Calculate the dynamic coupling coefficient based on the Reynolds number and structural flexibility coefficient: in, This represents the dynamic coupling coefficient.

[0029] In the above process, in the dynamic coupling coefficient In the calculation formula, the dependent variable Used to characterize the intensity of nonlinear interaction between fluid and structure in a semi-floating state. A large value indicates a significant fluid-structure interaction effect, and the wave load amplifies the stress on the structure. A smaller value means a weaker fluid-structure interaction effect, and a linear transfer function can more accurately describe the load-stress transfer relationship. by The item reflects the influence of the fluid flow state. When it increases As the coupling coefficient increases, it grows slowly at a logarithmic rate, reflecting the nonlinear enhancement effect of the change in flow state from laminar to turbulent transition zone on load transfer. The item reflects the influence of structural flexibility. The coupling coefficient increases linearly with increasing coefficient, reflecting the physical law that the more flexible the structure, the more significant the fluid-structure interaction effect and the stronger the reaction of structural deformation to fluid load. The rationality of this formula lies in the use of a logarithmic-linear combination to represent the fluid effect (…). Item) and structural effects ( The terms are listed separately, with the fluid term presented in logarithmic form to avoid [further issues]. exist When the magnitude changes, the coupling coefficient fluctuates drastically. The structural term adopts a linear form to reflect the direct proportional relationship between flexibility and coupling strength. The coefficients 0.28 and 1.5 are determined through dimensional compatibility and comparative analysis of hydrodynamic and structural responses under semi-floating conditions, making... Within the typical floating platform parameter range ( for , for ) produces approximately The correction magnitude is sufficient to capture the nonlinear coupling effect in the semi-floating state without excessively amplifying the reference response of the linear transfer function, thus providing the stress frequency domain response parameters in step S3. The calculation provides physically reasonable dynamic correction coefficients. From the perspective of positive and negative correlations, and Positive correlation, and They show a strict positive correlation, while and and It is negatively correlated, therefore Ultimately related to the material's elastic modulus and plate thickness Negatively correlated with column diameter They are positively correlated. Physically, the more rigid the structure, the smaller the coupling correction; the more flexible the structure or the stronger the fluid load, the larger the coupling correction. This is consistent with the actual physical laws of fluid-structure interaction in a semi-floating state.

[0030] In the above embodiments, 20 sets of data on the Reynolds number and the corresponding dynamic coupling coefficient are given to reflect the change of the dynamic coupling coefficient with the Reynolds number, as shown in Table 1: Table 1: Relationship between Reynolds number and corresponding dynamic coupling coefficient Table 1 above provides... Reynolds number from Increase to The dynamic coupling coefficient increased from 5.873 to 7.597. Follow It exhibits a logarithmic growth trend, with the growth rate increasing accordingly. Increases and gradually slows down, constant term This constitutes the baseline value of the coupling coefficient, reflecting the continuous contribution of the structure's own flexibility to the fluid-structure interaction effect in the semi-floating state.

[0031] S3: Perform Discrete Fourier Transform on the wave load time history data, interpolate the initial frequency response parameters to make them consistent with the frequency points of the wave load time history data after Discrete Fourier Transform, calculate the stress frequency domain response parameters based on the dynamic coupling coefficient and the interpolated frequency response parameters, and then perform Inverse Discrete Fourier Transform on the stress frequency domain response parameters to obtain the corresponding time domain stress time history parameters. The interpolated frequency response parameters are calculated as follows: Perform Discrete Fourier Transform on wave load time history data: in, In the program sequence representing the discrete wave load output by the wave load model, the first... Load values ​​at each time point; This indicates the frequency at which the discrete Fourier transform is applied. Wave load frequency domain data at the location; Indicates the first A discrete angular frequency point; Indicates a discrete time point index; Indicates the total length of the program sequence when dealing with wave loads; Indicates the first The time values ​​corresponding to each discrete time point; Indicates the index of discrete angular frequency points; The initial frequency response parameters are interpolated to match... By keeping the frequency points consistent, the interpolated frequency response parameters are obtained: for For each target frequency point, first determine whether the target frequency point is within the known frequency range of the preliminary frequency response parameters. If the target frequency point is lower than the known lowest frequency, then directly take the parameter value corresponding to the known lowest frequency point as the interpolation result of the target frequency point; if the target frequency point is higher than the known highest frequency, then directly take the parameter value corresponding to the known highest frequency point as the interpolation result of the target frequency point. If the target frequency point is within the known frequency range, then search from left to right in the sequence of known frequency points to find the first known frequency point that is greater than or equal to the target frequency point, and simultaneously take the previous known frequency point. These two points are the two known frequency points adjacent to the target frequency point. Then calculate the distance from the target frequency point to the previous known frequency point, and the total distance between the two known frequency points. The interpolated parameter value of the target frequency point can be obtained by adding the parameter value of the previous known frequency point to the change in the parameter value over the total distance and multiplying it by the distance ratio. The distance ratio is equal to the distance from the target frequency point to the previous known frequency point divided by the distance between the two known frequency points.

[0032] In the above process, the program sequence is obtained by processing the discrete wave load output by the wave load model. Perform a discrete Fourier transform to convert the data from the time domain to the frequency domain to obtain the wave load frequency domain data. The purpose is to decompose the time-domain payload sequence into its angular frequencies. The corresponding amplitude and phase components enable subsequent stress response calculations to be efficiently performed in the frequency domain via multiplication rather than convolution in the time domain. Since the preliminary frequency response parameters are derived from finite element harmonic response analysis, their known frequency points are determined by the unit harmonic excitation frequency set during the harmonic response analysis, and are typically related to... Discrete angular frequency points The inconsistencies necessitate interpolation to map the initial frequency response parameters to... For each target frequency point, extrapolation is used for target points outside the known frequency range, i.e., the lowest value is taken if the frequency is lower than the lowest frequency and the highest value is taken if the frequency is higher than the highest frequency. For target points within the known frequency range, linear interpolation is used, i.e., first locate the two known frequency points adjacent to the target frequency point, then use the Euclidean distance from the target frequency point to the previous point divided by the total distance between the two points as the weight, and linearly weight the parameter values ​​of the two points according to this weight to obtain the interpolation result. The technical effect achieved by this processing is that linear interpolation provides a first-order continuous frequency response estimate within the unknown frequency band, ensuring the smoothness of the frequency response parameters as they change with frequency, while avoiding the Runge phenomenon (i.e., the violent oscillation of high-order polynomials near the endpoints of the interval) that occurs with high-order interpolation when frequency points are sparse. Simultaneously, the extrapolation assignment ensures... All frequency points in the equation have corresponding frequency response parameter values, which makes the subsequent stress frequency domain response parameters... It can perform point-by-point calculations across the entire frequency band, thereby matching and fusing the finite element harmonic response analysis results with the wave load frequency domain data predicted by LSTM in the frequency domain, providing a complete frequency domain input for the final inverse discrete Fourier transform reconstruction of the time domain stress history.

[0033] The time-domain stress time history parameters are calculated as follows: Based on the dynamic coupling coefficient and the interpolated frequency response parameters, the stress frequency domain response parameters are calculated: in, Indicates at angular frequency Stress frequency domain response parameters at the location; This represents the frequency response parameter after interpolation; In the above process, the stress frequency domain response parameters In the calculation formula, the dependent variable Indicates at angular frequency The frequency domain stress response of the structure after fluid-structure interaction correction has a large value, indicating that the wave load at that frequency component contributes significantly to the stress of the structural hotspots, while a small value indicates that the excitation of that frequency component is attenuated during structural transmission or that the load energy at that frequency is low; the independent variable in this formula... Directly affected by multiplication The two are strictly positively correlated. The larger the value, the greater the stress response at the corresponding frequency, because wave load is the direct driving force of the structural response; independent variable Similarly, the effect is achieved through multiplication. This reflects the inherent dynamic transmission characteristics of the structure under a unit excitation. A large value implies that the structure exhibits a resonant amplification effect at that frequency, significantly enhancing the stress response; independent variable As a whole multiplicative factor, it affects all frequency components. A value greater than 1 indicates that the fluid-structure interaction nonlinearity in the semi-floating state amplifies the structural response. A value less than 1 indicates that the coupling effect weakens the transmission efficiency; The rationale for this formula lies in the fact that, under the assumption of a linear time-invariant system, the structural stress response is equal to the pointwise product of the load input and the system transfer function in the frequency domain. The introduction of this method modifies the traditional linear transitivity into a quasi-linear relationship that takes into account the nonlinear coupling effect of the semi-floating state, thus making... , and The physical factors represented by each of the three (fluid-structure interaction strength, wave load frequency distribution, and inherent dynamic characteristics of the structure) can be simply fused in the frequency domain in a multiplicative manner, avoiding the complex convolution calculations required to handle coupled nonlinear terms in the time domain.

[0034] Performing an inverse discrete Fourier transform on the stress frequency domain response yields the corresponding time-domain stress time-history parameters: in, After the inverse discrete Fourier transform, at the... Time-domain stress time history parameters at discrete time points.

[0035] In the above process, the dependent variable This indicates the result after the inverse discrete Fourier transform at the th... The time-domain stress time history parameter at each discrete time point indicates that the structure is experiencing a higher stress level at that moment, while a smaller value indicates a lower stress level. The independent variable in this formula is the frequency point. Stress frequency domain response parameters at the point Each Through complex exponents With time Correlation, all frequency components in time The stress value at that moment is reconstructed by coherent superposition of the components, and each frequency component is used to... The contribution depends on its magnitude and phase. The alignment at a certain point—when the phases of multiple frequency components tend to be consistent at a certain moment, the stress peak at that moment increases significantly, forming a stress peak in the time domain; the rationality of this formula lies in the fact that the inverse discrete Fourier transform is the strict mathematical inverse operation of the discrete Fourier transform, which can accurately recover any complex spectrum in the frequency domain into a real sequence in the time domain, that is, the imaginary parts cancel each other out in numerical calculations. The stress frequency domain response parameters are obtained by multiplying the frequency domain points by points through this transformation. Transforming back to the time domain yields the final time-domain stress time history. It retains the high efficiency of frequency domain computation, requiring only... Complex multiplication and addition, rather than temporal convolution. This operation provides the necessary stress-time history data for extracting time-domain stress cycle features, enabling subsequent fatigue damage assessment using the rainflow counting method, thus achieving a balance between frequency domain efficiency and time-domain accuracy. From the perspective of positive and negative correlations... Amplitude of each frequency component The overall correlation is positive, and the larger the amplitude of the frequency component, the more significant its contribution to the time-domain stress. However, the stress value at a specific time point also depends on the relative phase relationship of each frequency component. Therefore, even if the amplitude of each frequency point remains unchanged, changing the phase distribution will change the waveform and peak distribution of the time-domain stress time history.

[0036] S4: Construct a time-varying corrosion factor based on the material parameters, use the rainflow counting method to extract the stress amplitude and stress cycle number from the time-domain stress time history parameters, and use the time-varying corrosion factor as a correction term to construct the corrected stress cycle number; calculate the cumulative damage based on the stress cycle number, and issue an alarm when the cumulative damage is greater than or equal to the preset damage value.

[0037] The cumulative damage is calculated as follows: Based on material parameters, a time-varying corrosion factor is constructed: in, express The time-varying corrosion factor at any given moment; Indicates the corrosion rate of the floating platform material; Represents a time variable; In the above process, under the time-varying corrosion factor In the middle, the dependent variable express The degree of fatigue performance degradation of a material over time, with a value between 1 and 0, varies with service time. Monotonically decreasing, A value close to 1 indicates that the material's fatigue performance has not significantly degraded in the early stages of corrosion. A value approaching 0 indicates a severe decrease in the material's fatigue resistance during the later stages of corrosion; the independent variable is the corrosion rate. and time variables All Negative correlation The larger or The longer The smaller the value, the better. This is because when floating platforms are in long-term service in the marine environment, chloride ions and dissolved oxygen in the seawater continuously erode the surface of the structure, leading to material cross-section loss and the initiation of surface microcracks. Moreover, corrosion damage accumulates exponentially over time. That is, when the protective layer is still effective in the early stage of corrosion, the degradation is slow. As corrosion products peel off and the protective layer is destroyed, the degradation rate gradually accelerates and then tends to stabilize due to the coverage of corrosion products. The exponential decay form can just describe this actual corrosion degradation process that is slow at first, then fast, and then slows down again. Time variable In the formula, it is used as a continuous time variable to characterize the total service time of the floating platform from the start of its service to the present moment. When the rainflow counting method is used to analyze the stress time history in the time domain... Extract the first Amplitude of secondary stress cycles At that time, the time interval corresponding to this stress cycle starts from the moment of the initial trough of the cycle. To the peak time However, due to time-varying corrosion factors This describes the long-term degradation effect of accumulated corrosion damage in structural materials throughout their service life, rather than instantaneous changes associated with a single stress cycle timescale. The value should be uniformly taken as the end time of the total service period corresponding to the entire time domain stress time history analyzed when counting rainflows, that is, the cumulative service life corresponding to the acquisition time of the stress time history data of that segment. The stress amplitude and stress cycle number were extracted from the time-domain stress time history parameters using the rainflow counting method, and the time-varying corrosion factor was used as a correction term to construct the corrected SN curve. in, Indicates the first The amplitude of the second stress cycle; Indicates the first Secondary stress cycle amplitude The number of stress cycles that a floating platform can withstand under the action of force; The slope of the SN curve represents the floating platform. This represents the intercept constant of the SN curve for a floating platform; In the above process, in the modified SN curve formula In the middle, the dependent variable Indicates the first Secondary stress cycle amplitude The number of stress cycles a floating platform can withstand under stress is a key factor. A larger value indicates stronger fatigue resistance and longer lifespan under that stress amplitude, while a smaller value indicates a higher likelihood of fatigue failure under that stress amplitude. The stress amplitude in this formula... by The reciprocal form of the power affects The two are strongly negatively correlated, meaning that as the stress amplitude increases, the fatigue life decreases sharply according to a power function relationship. This reflects the double logarithmic linear relationship of the SN curve in Miner's fatigue theory. ), A larger value indicates a more sensitive structure to changes in stress amplitude, consistent with the physical laws of fatigue damage in welded steel structures in marine environments; time-varying corrosion factor As a multiplicative correction term, it acts on the intercept constant of the SN curve. , Decrease A proportional decrease indicates that corrosion reduces the overall fatigue life of the structure under the same stress amplitude. Physically, this is equivalent to corrosion weakening the effective cross-section of the structure and increasing the local stress concentration factor, thus affecting the SN curve over service time. translate downwards; The rationale behind this formula lies in the following: and These are the two intrinsic parameters of the SN curve ( The intercept constant is... (slope), time-varying corrosion factor Only corrections Without change This is because corrosion primarily weakens the load-bearing area of ​​the structure and initiates cracks, thus reducing fatigue life under the same stress amplitude, but increasing the stress sensitivity of crack propagation (i.e., Due to the inherent properties of the material, it is relatively less affected by corrosion.

[0038] Calculate cumulative damage based on the number of stress cycles: in, Indicates cumulative damage; This represents the total number of stress cycle sets extracted by the rainflow counting method; Indicates the stress cycle number index; In the above process, the dependent variable This represents the total cumulative damage to the structure after experiencing M sets of stress cycles. A large value indicates that the structure has consumed a significant portion of its fatigue life reserve and is nearing failure. A smaller value means the structure is still within a safe service range; in this formula, the first... The actual number of stress cycles occurring in the group and A positive correlation exists, meaning the more actual number of cycles, the greater the cumulative damage; while the allowable number of cycles corresponds to this set of stress amplitudes. and Negative correlation A smaller value indicates that the structure can withstand fewer cycles at that stress amplitude, thus reducing the damage caused by each cycle. The larger; It itself is obtained by modifying the SN curve. and Joint decision: The larger The smaller, The smaller The smaller they are, the more they decrease. Indirectly increases cumulative damage The rationale behind this formula lies in the fact that the Miner linear accumulation criterion superimposes fatigue damage under different stress amplitudes with equal weights. This method is simple in form and has been proven effective through extensive marine engineering fatigue analysis practice. Time-varying corrosion factors have been embedded in it. The correction made This reflects the continuous effect of corrosion over time. The larger The smaller, Overall decrease, cumulative damage over the same number of cycles This increases the capacity, thereby enabling time-varying assessment of structural fatigue safety status throughout the entire life cycle within the early warning mechanism; The slope of the SN curve and the intercept constant of the SN curve Based on the material type, connection method, and environmental conditions of the key components of the floating platform, the relevant classification society specifications or international standards such as DNV GL-RP-C203, IIW (International Institute of Welding Recommended Code for Fatigue Design or ABS "Guidelines for Fatigue Assessment of Ships and Offshore Structures") are consulted to determine the following: For DH36 / EH36 high-strength steel and its welded joints commonly used in marine engineering, It is usually taken as 3. or 3.5 to 4.0, while The fatigue strength is determined by the specific structural detail classification level. Different detail levels correspond to different fatigue strengths. For example, for T-type or K-type welded joints similar to those connecting to columns and pontoons, the corresponding curve level (such as F2, G, or W3) in the specification can be found. value.

[0039] When the accumulated damage is greater than or equal to the preset damage value, it indicates that the floating platform is in a safe state; otherwise, it indicates that the floating platform is in a dangerous state and an alarm will be issued immediately.

[0040] In the above process, the preset damage value is determined based on the target reliability or allowable cumulative damage threshold corresponding to the design fatigue life specified in the floating platform design specifications. According to classification society standards such as DNV GL-RP-C203 or ABS "Guidelines for Fatigue Assessment of Ships and Offshore Structures," when using the SN curve and Miner linear cumulative criterion for fatigue assessment, the cumulative damage value of 1.0 corresponding to the design fatigue life is usually used as the benchmark threshold. However, since the semi-floating state only occupies a portion of the platform's entire lifespan and material corrosion causes a continuous decline in actual fatigue performance, a safety factor needs to be incorporated into the benchmark threshold for reduction when determining the preset damage value. Therefore, the preset damage value is set to [value missing]. ,in The safety factor, typically ranging from 2 to 5, ensures sufficient safety margin to cover wave load prediction errors, material parameter uncertainties, and statistical variability of fluid-structure interaction correction coefficients in a semi-floating state, provided that the cumulative damage does not exceed the reduction threshold. This ensures that the early warning mechanism issues an alert before the structure reaches its design life, allowing sufficient reaction time for operation and maintenance decisions.

[0041] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0042] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0043] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for assessing wave load damage to a semi-floating floating platform structure, characterized in that, include: S1: Obtain wave environment parameters of the target sea area, extract geometric and material parameters of key parts of the floating platform, establish a finite element model of the floating platform, and perform harmonic response analysis on the finite element model to obtain preliminary frequency response parameters. S2: Based on the wave environment parameters, construct a JONSWAP spectrum model and calculate the wave spectrum accordingly. Use the energy segmentation method to discretize the wave spectrum into multiple frequency points, construct a wave load model based on LSTM, and input the discretized wave spectrum into the trained wave load model to obtain wave load time history data. Based on the wave environment parameters, geometric parameters, and material parameters, calculate the Reynolds number and structural flexibility coefficient, and further calculate the dynamic coupling coefficient accordingly. S3: Perform Discrete Fourier Transform on the wave load time history data, interpolate the initial frequency response parameters to make them consistent with the frequency points of the wave load time history data after Discrete Fourier Transform, calculate the stress frequency domain response parameters based on the dynamic coupling coefficient and the interpolated frequency response parameters, and then perform Inverse Discrete Fourier Transform on the stress frequency domain response parameters to obtain the corresponding time domain stress time history parameters. S4: Construct a time-varying corrosion factor based on the material parameters, use the rainflow counting method to extract the stress amplitude and stress cycle number from the time-domain stress time history parameters, and use the time-varying corrosion factor as a correction term to construct the corrected stress cycle number; The cumulative damage is calculated based on the number of stress cycles, and an alarm is issued when the cumulative damage is greater than or equal to a preset damage value.

2. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 1, characterized in that, Obtain wave spectrum data for the target sea area, and extract the geometric and material parameters of the floating platform, specifically: Wave environment parameters, including significant wave height and spectral peak period, are obtained from a marine meteorological database. At the same time, geometric and material parameters of key parts of the floating platform are also obtained. The key parts are the connection between the column and the pontoon. The geometric parameters include the column diameter and plate thickness, and the material parameters include the elastic modulus, Poisson's ratio, and corrosion rate. A finite element model of the floating platform is established, and harmonic response analysis is performed on the finite element model. Specifically, a unit harmonic excitation along the wave propagation direction is applied to the floating platform, the stress response amplitude of the key parts at each frequency point of the unit harmonic excitation is extracted, and the ratio of the stress response amplitude to the excitation amplitude is calculated to obtain preliminary frequency response parameters.

3. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 2, characterized in that, Constructing the JONSWAP spectral model and calculating the wave spectrum based on it, specifically: Based on wave environment parameters, construct the JONSWAP spectral model: in, Indicates the peak factor; Indicates the significant wave height; Indicates the period of the spectral peak; Indicates angular frequency; Indicates angular frequency as Wave spectrum at time; This represents the peak shape parameter.

4. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 3, characterized in that, The wave spectrum is discretized into multiple frequency points using the energy segmentation method, specifically as follows: The total wave energy is calculated as follows: in, This represents the total wave energy; Indicates the angular frequency of the spectral peak; Indicates the discrete lower limit frequency; Indicates the discrete upper limit frequency; Divide the total wave energy into equal parts to obtain the unit energy value: in, Indicates the value per unit of energy; Indicates the number of energy portions; Solve for the boundary values ​​at each frequency point: in, Indicates the energy boundary frequency; Indicates the index of the number of energy segments; The final discrete frequency point set is: ,in, It represents a set of discrete frequencies.

5. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 4, characterized in that, The specific logic for obtaining the wave load time history data is as follows: Multiple sets of discrete wave spectra and corresponding discrete wave load time sequence data are obtained from the HELOFOW project database. The discrete wave spectra are used as the input of the wave load model, and the discrete wave load time sequence data are used as the output of the wave load model. The wave load model is trained. When the error between the predicted output result of consecutive preset training rounds and the true value of the discrete wave load time sequence data does not exceed the set prediction error, the training is considered complete. The wave spectrum, discretized into multiple frequency points, is input into the trained wave load model to obtain the corresponding discrete wave load time program. The wave load model is based on an LSTM model architecture, specifically including an input layer, a first LSTM layer, a second LSTM layer, a third LSTM layer, a first dropout layer, a second dropout layer, a third dropout layer, a fully connected layer, and an output layer. The input data for the input layer is a wave spectrum discretized into multiple frequency points, and the output data of the input layer is input to the first LSTM layer. The first LSTM layer is a first Long Short-Term Memory (LSTM) network layer containing 128 memory units, and the output data of the first LSTM layer is input to both the second LSTM layer and the first dropout layer. The second LSTM layer is a second LSTM layer containing 128 memory units. The third LSTM layer is a Long Short-Term Memory (LSTM) network layer. The output data of the second LSTM layer and the output data of the first discard layer are input into the third LSTM layer. The third LSTM layer is a third LSTM network layer containing 128 memory units. The output data of the third LSTM layer is input into the second discard layer. The output data of the second discard layer is input into the fully connected layer. The output data of the fully connected layer is input into the third discard layer. The output data of the third discard layer is input into the output layer. The output data of the output layer is a discrete wave load time sequence. The discard rate of the first, second, and third discard layers is 0.

2.

6. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 2, characterized in that, The dynamic coupling coefficient is calculated as follows: Calculate the Reynolds number and structural compliance coefficient based on wave environment parameters, geometric parameters, and material parameters: in, Indicates the diameter of the column; Indicates the density of seawater; Indicates the significant wave height; Indicates the period of the spectral peak; Represents the Reynolds number; Indicates the elastic modulus of the material in critical components; Indicates the plate thickness of key components; Indicates the Poisson's ratio of the material in the critical component; Indicates the structural flexibility coefficient; Indicates the dynamic viscosity of seawater; Represents gravitational acceleration; Calculate the dynamic coupling coefficient based on the Reynolds number and structural flexibility coefficient: in, This represents the dynamic coupling coefficient.

7. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 1, characterized in that, The interpolated frequency response parameters are calculated as follows: Perform Discrete Fourier Transform on wave load time history data: in, In the program sequence representing the discrete wave load output by the wave load model, the first... Load values ​​at each time point; This indicates the frequency at which the discrete Fourier transform is applied. Wave load frequency domain data at the location; Indicates the first A discrete angular frequency point; Indicates a discrete time point index; Indicates the total length of the program sequence when dealing with wave loads; Indicates the first The time values ​​corresponding to each discrete time point; Indicates the index of discrete angular frequency points; The initial frequency response parameters are interpolated to match... By keeping the frequency points consistent, the interpolated frequency response parameters are obtained: for For each target frequency point, first determine whether the target frequency point is within the known frequency range of the preliminary frequency response parameters. If the target frequency point is lower than the known lowest frequency, then directly take the parameter value corresponding to the known lowest frequency point as the interpolation result of the target frequency point; if the target frequency point is higher than the known highest frequency, then directly take the parameter value corresponding to the known highest frequency point as the interpolation result of the target frequency point. If the target frequency point is within the known frequency range, then search from left to right in the sequence of known frequency points to find the first known frequency point that is greater than or equal to the target frequency point, and simultaneously take the previous known frequency point. These two points are the two known frequency points adjacent to the target frequency point. Then calculate the distance from the target frequency point to the previous known frequency point, and the total distance between the two known frequency points. The interpolated parameter value of the target frequency point can be obtained by adding the parameter value of the previous known frequency point to the change in the parameter value over the total distance and multiplying it by the distance ratio. The distance ratio is equal to the distance from the target frequency point to the previous known frequency point divided by the distance between the two known frequency points.

8. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 7, characterized in that, The time-domain stress time history parameters are calculated as follows: Based on the dynamic coupling coefficient and the interpolated frequency response parameters, the stress frequency domain response parameters are calculated: in, Indicates at angular frequency Stress frequency domain response parameters at the location; This represents the frequency response parameter after interpolation; Performing an inverse discrete Fourier transform on the stress frequency domain response yields the corresponding time-domain stress time-history parameters: in, After the inverse discrete Fourier transform, at the... Time-domain stress time history parameters at discrete time points.

9. The method for assessing wave load damage to a semi-floating floating platform structure according to claim 8, characterized in that, The cumulative damage is calculated as follows: Based on material parameters, a time-varying corrosion factor is constructed: in, express The time-varying corrosion factor at any given moment; Indicates the corrosion rate of the floating platform material; Represents a time variable; The stress amplitude and stress cycle number were extracted from the time-domain stress time history parameters using the rainflow counting method, and the time-varying corrosion factor was used as a correction term to construct the corrected SN curve. in, Indicates the first The amplitude of the second stress cycle; Indicates the first Secondary stress cycle amplitude The number of stress cycles that a floating platform can withstand under the action of force; The slope of the SN curve represents the floating platform. This represents the intercept constant of the SN curve for a floating platform. Calculate cumulative damage based on the number of stress cycles: in, Indicates cumulative damage; This represents the total number of stress cycle sets extracted by the rainflow counting method; Indicates the stress cycle number index; When the accumulated damage is greater than or equal to the preset damage value, it indicates that the floating platform is in a safe state; otherwise, it indicates that the floating platform is in a dangerous state and an alarm will be issued immediately.

Images