Semi-submersible platform air gap extremum forecasting system and method based on numerical simulation

Abstract

The invention provides a semi-submersible platform air gap extremum forecasting system and method based on numerical simulation, and relates to the technical field of numerical simulation, and the method comprises the steps: obtaining an air gap response sequence of a semi-submersible platform in a target water area through numerical simulation; extracting low-frequency response from the air gap response sequence, and performing nonlinear correction on the low-frequency response based on a data driving model to obtain a corrected low-frequency response sequence; reconstructing the corrected low-frequency response sequence and the wave frequency response to obtain a composite air gap response sequence; performing non-Gaussian extremum statistics based on the composite air gap response sequence to obtain an air gap extremum prediction result; according to the method and the device, the test sequence of the air gap of the semi-submersible platform along with the time change is obtained, the data driving model is iteratively optimized on the basis of the difference between the test sequence and the air gap extreme value prediction result, and the data driving type accurate correction aiming at the low and medium frequency response in numerical simulation can be realized, so that the air gap extreme value prediction accuracy of the semi-submersible platform is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of numerical simulation, more particularly, the present application relates to a semi-submersible platform air gap extreme value prediction system and method based on numerical simulation. BACKGROUND

[0002] Numerical simulation technology is the core means of current semi-submersible platform performance evaluation and safety prediction. By establishing a hydrodynamic model of the semi-submersible platform and solving the fluid dynamics control equation, the motion and load response of the platform in complex water areas are simulated to provide key data support for design optimization and safe operation. In the design and operation safety analysis of semi-submersible platforms, air gap extreme value prediction is crucial. A small air gap can cause severe impact of waves on the deck, seriously threatening the structural safety of the semi-submersible platform and the integrity of personnel and equipment.

[0003] Developing a high-precision air gap extreme value prediction is of great significance to ensuring the safety of the platform throughout its life cycle. Currently, numerical simulation based on the frequency domain or time domain is the mainstream method for air gap extreme value prediction. By calculating the relative relationship between the platform motion and wave elevation caused by waves, the statistical characteristics and extreme values of the air gap are predicted. However, the existing prediction methods still have obvious technical limitations. When dealing with low-frequency motion caused by second-order difference frequency forces, due to the simplification of complex physical phenomena such as viscous damping and vortex-induced effects, there is a systematic and inherent bias in the prediction of low-frequency response that cannot be completely eliminated by adjusting parameters. Moreover, there is a lack of understanding of the main sources of error. Errors in simulation are directly passed on and amplified in the final extreme value prediction results, making it difficult to effectively feedback test data to the numerical model to achieve automatic calibration of model parameters and continuous iterative optimization of prediction capability. Therefore, how to achieve data-driven precise correction of low-frequency response in numerical simulation to improve the accuracy of air gap extreme value prediction of semi-submersible platforms is a difficult problem faced by the industry. SUMMARY

[0004] The present application provides a semi-submersible platform air gap extreme value prediction system and method based on numerical simulation, which can achieve data-driven precise correction of low-frequency response in numerical simulation to improve the accuracy of air gap extreme value prediction of semi-submersible platforms.

[0005] In a first aspect, the present application provides a semi-submersible platform air gap extreme value prediction method based on numerical simulation, the prediction method comprising the following steps:

[0006] Obtaining an air gap response sequence of the semi-submersible platform in a target water area through numerical simulation, the air gap response sequence including wave frequency response and low-frequency response;

[0007] Extracting the low-frequency response from the air gap response sequence, and performing nonlinear correction on the low-frequency response based on a data-driven model to obtain a corrected low-frequency response sequence.

[0008] The modified low-frequency response sequence and the wave frequency response are reconstructed to obtain the composite air gap response sequence.

[0009] Based on the composite air gap response sequence, non-Gaussian extreme value statistics are performed to obtain the air gap extreme value prediction results for the target water area.

[0010] A test sequence of air gap variation over time for a semi-submersible platform was obtained through wave pool tests. The data-driven model was iteratively optimized based on the difference between the test sequence and the predicted extreme values ​​of the air gap.

[0011] In this embodiment, obtaining the air gap response sequence of the semi-submersible platform in the target water area through numerical simulation specifically includes:

[0012] The transfer functions of motion and wave load of the semi-submersible platform are constructed based on the structural parameters of the semi-submersible platform and the wave parameters of the target water area.

[0013] The wave frequency motion response and wave load are determined by the transfer function.

[0014] The air gap response sequence is obtained by performing coupled analysis on the wave frequency motion response and wave load.

[0015] In this embodiment, the low-frequency response is extracted from the air gap response sequence based on a high-pass filter.

[0016] In this embodiment, the data-driven model is a machine learning model based on historical data, which is used to learn and correct the inherent low-frequency response deviation in the numerical simulation through the modal features of the low-frequency response.

[0017] In this embodiment, the low-frequency response is nonlinearly corrected based on a data-driven model to obtain the corrected low-frequency response sequence, which specifically includes:

[0018] Modal decomposition of the low-frequency response yields multiple intrinsic mode function components;

[0019] Construct a training set based on all intrinsic mode function components;

[0020] All intrinsic mode function components are input into the data-driven model trained on the training set for signal reconstruction, resulting in the corrected low-frequency response sequence.

[0021] In this embodiment, reconstructing the modified low-frequency response sequence and the wave frequency response to obtain the composite air gap response sequence specifically includes:

[0022] The modified low-frequency response sequence and the wave frequency response are linearly superimposed in the time domain to obtain the initial reconstructed sequence;

[0023] The initial reconstructed sequence is subjected to phase consistency testing and amplitude normalization to obtain the composite air gap response sequence.

[0024] In this embodiment, the non-Gaussian extreme value statistics based on the composite air gap response sequence are used to obtain the air gap extreme value prediction result for the target water area, specifically including:

[0025] Higher-order statistical analysis was performed on the composite air gap response sequence to obtain skewness and kurtosis indices.

[0026] A polynomial transformation model based on non-Gaussian properties is constructed based on the skewness index and the kurtosis index.

[0027] Monte Carlo simulation is applied to the polynomial transformation model to generate a synthetic sequence, and then peak over-threshold sampling is performed on the synthetic sequence to obtain extreme value samples;

[0028] The extreme value samples are fitted using a generalized Pareto distribution to obtain an extreme value distribution function. Then, feature values ​​corresponding to a preset return period are extracted from the extreme value distribution function to obtain the air gap extreme value prediction result for the target water area.

[0029] In this embodiment, the test sequence for obtaining the time-varying air gap of a semi-submersible platform through wave pool testing specifically includes:

[0030] Construct a physical model of the semi-submersible platform;

[0031] A wave field is generated based on the wave parameters of the target water area;

[0032] The physical model is placed in the wave field, and the wave surface rise around the semi-submersible platform is measured by a wave height sensor to obtain an experimental sequence of the air gap of the semi-submersible platform changing over time.

[0033] In this embodiment, iterative optimization of the data-driven model based on the difference between the experimental sequence and the air gap extreme value prediction results specifically includes:

[0034] The response residual sequence is determined using the experimental sequence and the air gap extreme value prediction results;

[0035] A loss function is constructed based on the response residual sequence, and the parameter gradient of the data-driven model is calculated using the gradient descent algorithm.

[0036] The internal weight parameters of the data-driven model are iteratively updated according to the gradient of the parameters until the loss function converges, thus completing the iterative optimization of the data-driven model.

[0037] Secondly, this application provides a numerical simulation-based air gap extreme value prediction system for semi-submersible platforms, used to execute a numerical simulation-based air gap extreme value prediction method for semi-submersible platforms, the prediction system comprising:

[0038] The numerical simulation module is used to obtain the air gap response sequence of the semi-submersible platform in the target water area through numerical simulation. The air gap response sequence includes wave frequency response and low frequency response.

[0039] The low-frequency correction module is used to extract the low-frequency response from the air gap response sequence, and perform nonlinear correction on the low-frequency response based on a data-driven model to obtain a corrected low-frequency response sequence.

[0040] The response reconstruction module is used to reconstruct the corrected low-frequency response sequence and the wave frequency response to obtain a composite air gap response sequence.

[0041] The extreme value prediction module is used to perform non-Gaussian extreme value statistics based on the composite air gap response sequence to obtain the air gap extreme value prediction result of the target water area;

[0042] The model optimization module is used to obtain a test sequence of the air gap of a semi-submersible platform changing over time through wave pool tests, and to iteratively optimize the data-driven model based on the difference between the test sequence and the predicted extreme values ​​of the air gap.

[0043] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0044] The air gap response sequence of a semi-submersible platform in the target water area is obtained through numerical simulation. The air gap response sequence includes wave frequency response and low-frequency response. The low-frequency response is extracted from the air gap response sequence, and nonlinear correction is performed on the low-frequency response based on a data-driven model to obtain a corrected low-frequency response sequence. The corrected low-frequency response sequence and the wave frequency response are reconstructed to obtain a composite air gap response sequence. Non-Gaussian extreme value statistics are performed on the composite air gap response sequence to obtain the air gap extreme value prediction result for the target water area. The test sequence of the air gap of the semi-submersible platform changing over time is obtained through wave pool tests. The data-driven model is iteratively optimized based on the difference between the test sequence and the air gap extreme value prediction result.

[0045] Therefore, this application can achieve data-driven precise correction of low-frequency response in numerical simulation. First, an air gap response sequence containing both wave frequency and low-frequency response is obtained through numerical simulation, establishing a benchmark data model covering the dynamic characteristics of the entire frequency band, providing a complete analytical foundation for subsequent precise correction. Second, the low-frequency response is extracted from the air gap response sequence, and nonlinear correction is performed on the low-frequency response based on the data-driven model, which is beneficial for precise intervention in numerical simulation errors. Through the nonlinear mapping capability of the data-driven model, the inherent bias of potential flow theory in simulating low-frequency motion can be effectively compensated, which is beneficial for improving prediction accuracy. Then, the corrected low-frequency response sequence is reconstructed with the more reliable wave frequency response to obtain a composite air gap response sequence, which can be used to correct the calibrated low-frequency response. The high-frequency components are organically integrated with the undisturbed high-frequency components to generate a more accurate time-domain response across the entire frequency band, which in turn provides high-quality input data for extreme value statistics. Next, non-Gaussian extreme value statistics are performed based on the high-precision composite air gap response sequence. Due to the significantly improved reliability of the input data, the extreme value statistics can more realistically reflect the tail characteristics of the air gap response, thus outputting extreme value prediction results that are more consistent with engineering practice. Finally, a validation sequence is obtained through wave pool experiments, and the data-driven model is iteratively optimized based on the difference between the validation sequence and the prediction results. This allows for the construction of a complete evolutionary closed-loop mechanism, continuously optimizing and correcting the data-driven model using the most reliable experimental data. This enables the data-driven model to have the ability to generalize to new water areas, thereby ensuring that the air gap extreme value prediction remains highly accurate.

[0046] In summary, the technical solution adopted in this application can achieve data-driven precise correction of low-frequency response in numerical simulation, thereby improving the accuracy of air gap extreme value prediction for semi-submersible platforms. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a flowchart of a method for predicting air gap extreme values ​​of a semi-submersible platform based on numerical simulation, provided in this application.

[0049] Figure 2 This is an exemplary flowchart of determining the modified low-frequency response sequence according to the present application;

[0050] Figure 3 This is an exemplary flowchart based on the prediction results of the air gap extreme value of the target water area provided in this application;

[0051] Figure 4 This is a module structure diagram of a semi-submersible platform air gap extreme value prediction system based on numerical simulation provided in this application. Detailed Implementation

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0053] This application provides a numerical simulation-based air gap extreme value prediction system and method for semi-submersible platforms. The core of this system involves obtaining the air gap response sequence of a semi-submersible platform in a target water area through numerical simulation. This air gap response sequence includes a wave frequency response and a low-frequency response. The low-frequency response is extracted from the air gap response sequence, and a nonlinear correction is applied to the low-frequency response based on a data-driven model to obtain a corrected low-frequency response sequence. The corrected low-frequency response sequence and the wave frequency response are reconstructed to obtain a composite air gap response sequence. Non-Gaussian extreme value statistics are performed on the composite air gap response sequence to obtain the predicted air gap extreme value for the target water area. A wave pool test is conducted to obtain a test sequence showing the air gap variation of the semi-submersible platform over time. The data-driven model is iteratively optimized based on the difference between the test sequence and the predicted air gap extreme value.

[0054] Example 1: To better understand the above technical solution, the following will provide a detailed description of the technical solution in conjunction with the accompanying drawings and specific implementation methods. (Refer to...) Figure 1 As shown in the figure, this is a flowchart of a method for predicting the air gap extreme value of a semi-submersible platform based on numerical simulation according to this embodiment of the present application. The prediction method includes the following steps:

[0055] In step S1, the air gap response sequence of the semi-submersible platform in the target water area is obtained through numerical simulation. The air gap response sequence includes wave frequency response and low frequency response.

[0056] In this embodiment, the air gap response sequence of the semi-submersible platform in the target water area can be obtained through numerical simulation in the following manner:

[0057] The transfer functions of motion and wave load of the semi-submersible platform are constructed based on the structural parameters of the semi-submersible platform and the wave parameters of the target water area.

[0058] The wave frequency motion response and wave load are determined by the transfer function.

[0059] The air gap response sequence is obtained by performing coupled analysis on the wave frequency motion response and wave load.

[0060] In specific implementation, firstly, the structural parameters, including the platform's main dimensions, column spacing, draft, and mass distribution, can be obtained from the platform design drawings; the wave parameters, including wave spectrum type, significant wave height, and peak period, can be obtained through long-term observation and statistics of the target water area. A wetted surface model of the platform is established using three-dimensional potential flow theory software, and the velocity potential under the action of regular waves of unit amplitude is solved using the boundary element method. Then, the complex response amplitudes of the platform's six degrees of freedom motion (i.e., sway, pitch, heave, roll, pitch, and yaw) and wave loads (wave forces and moments in each direction) relative to the wave frequency and direction are calculated. The matrix form of the above complex response amplitude data is used as the transfer function of the semi-submersible platform's motion and wave loads. Then, the wave parameters can be discretized into multiple regular wave components using the JONSWAP spectrum. For each regular wave component, the regular wave component's... Multiplying the wave amplitude by the complex response amplitude in the transfer function yields the complex form of the motion response and wave load of the semi-submersible platform under the regular wave component. This process is repeated to obtain the complex forms of the motion response and load of the semi-submersible platform under all regular wave components. Then, all complex forms are linearly superimposed in the time domain using an inverse fast Fourier transform. The resulting motion response is taken as the wave frequency motion response, and the resulting load is taken as the wave load. Finally, based on linear wave theory, the incident wavefront can be generated from the JONSWAP spectrum. Multiple air gap monitoring points are pre-set below the semi-submersible platform deck. Sensor monitoring technology is used to obtain the difference between the instantaneous wavefront rise (considering wave diffraction and radiation effects) and vertical motion (synthesized from heave, pitch, and roll in the wave frequency motion response) at the air gap monitoring points. The result of arranging all differences in chronological order is taken as the air gap response sequence.

[0061] It should be noted that, in this embodiment, the transfer function describes the linear relationship between the motion of the semi-submersible platform and the wave load it experiences in the frequency domain and the incident wave. The transfer function is essentially a complex matrix, where the modulus of the complex matrix represents the response amplitude and the argument of the complex matrix represents the phase lag. The wave frequency motion response specifically refers to the motion of the semi-submersible platform caused by waves in the wave frequency range (usually 0.05-2.0 rad / s), but does not include low-frequency motion caused by second-order differential frequency forces. The wave load refers to the total wave force and torque obtained by integrating the first-order and second-order pressures induced by the waves acting on the wetted surface of the platform. The air gap response sequence refers to the numerical sequence of the net height above the still water surface changing over time, where negative values ​​in the numerical sequence indicate that waves may impact the bottom of the deck.

[0062] In step S2, the low-frequency response is extracted from the air gap response sequence, and the low-frequency response is nonlinearly corrected based on the data-driven model to obtain the corrected low-frequency response sequence.

[0063] In this embodiment, the low-frequency response is extracted from the air gap response sequence using a high-pass filter. Specifically, the high-pass filter's technical parameters are first determined, with its cutoff frequency set to 0.04 Hz. This cutoff frequency is a key threshold for distinguishing between wave frequency and low-frequency responses; wave components below the cutoff frequency are typically excited by second-order difference-frequency forces. Then, a Butterworth filter is used to reduce phase information interference in the air gap response sequence. Finally, the high-pass filter filters the air gap response sequence, retaining the sequence containing components with frequencies below the cutoff frequency, which is then used as the low-frequency response.

[0064] It should be noted that, in this embodiment, the high-pass filter is an electronic or digital filtering device that allows high-frequency signals to pass through while suppressing or attenuating low-frequency signals. In this application, the high-pass filter can separate the sequence of components below the cutoff frequency in the air gap response sequence. The low-frequency response specifically refers to the air gap change component generated by the second-order difference frequency force of the wave, which usually has a frequency below the cutoff frequency. The low-frequency response is the main source of systematic errors in the prediction of air gap extreme values.

[0065] Preferably, in this embodiment, reference Figure 2 As shown, this figure is an exemplary flowchart of determining the corrected low-frequency response sequence according to the present application. In this embodiment, the low-frequency response is nonlinearly corrected based on a data-driven model to obtain the corrected low-frequency response sequence, which can be achieved by the following steps:

[0066] First, in step S21, the low-frequency response is decomposed into modes to obtain multiple intrinsic mode function components;

[0067] Then, in step S22, a training set is constructed based on all intrinsic mode function components;

[0068] Finally, in step S23, all intrinsic mode function components are input into the data-driven model trained by the training set for signal reconstruction to obtain the corrected low-frequency response sequence.

[0069] It should be noted that, in this embodiment, the data-driven model is a machine learning model based on historical data, used to learn and correct the inherent low-frequency response deviation in the numerical simulation through the modal features of the low-frequency response. Preferably, the machine learning model can use a long short-term memory neural network model, which can be used to establish a mapping relationship between the erroneous modal features and the corrected values ​​that are closer to the real state. Here, historical data refers to a set of data pairs that are accumulated in advance and contain multiple sets of matching "low-frequency responses obtained from numerical simulation" and "corresponding low-frequency responses obtained through physical model experiments or high-precision sensor measurements". This set of data pairs can constitute a knowledge base for training the machine learning model and is the basis for it to learn and capture the simulation deviation patterns.

[0070] In practical implementation, firstly, a variational mode decomposition algorithm can be used for processing. For example, setting the number of intrinsic mode function components K to 6 and the penalty parameter α to 2000, the low-frequency response is decomposed into 6 modal components arranged from low to high frequency by iteratively searching for the center frequency and signal distribution that minimizes the sum of the estimated bandwidths of each mode. These 6 modal components are then used as intrinsic mode function components. Next, all intrinsic mode function components extracted from historical numerical simulation cases can be used as input features, and the intrinsic mode function components obtained after the low-frequency response undergoes variational mode decomposition with the same parameters can be used as the target. The output is randomly divided into a training subset and a test subset in a 7:3 ratio. The combination of the training subset and the test subset is used as the training set. Finally, the long short-term memory neural network model is trained using the training set containing the training subset and the test subset. All intrinsic mode function components are input into the trained long short-term memory neural network model in ascending order of frequency. The long short-term memory neural network model outputs all the corrected mode components. All the corrected mode components are linearly superimposed in the time domain, and the resulting sequence is used as the corrected low-frequency response sequence.

[0071] It should be noted that, in this embodiment, intrinsic mode function components refer to quasi-orthogonal sub-signals with different center frequencies. Each intrinsic mode function component represents the oscillation mode of a specific frequency band in the original signal, which can decompose complex mixed signals into multiple relatively simple sub-signals. The training set in this application is a labeled data set used to train the data-driven model, which can provide learning samples and enable the data-driven model to master the mapping law from the simulation mode with errors to the real mode. Signal reconstruction refers to the process of re-merging the various intrinsic mode function components after correction by the data-driven model into a complete time-domain signal. Signal reconstruction is the inverse operation of mode decomposition and can fuse the corrected frequency band signal components into a complete and corrected low-frequency response sequence.

[0072] In step S3, the corrected low-frequency response sequence and the wave frequency response are reconstructed to obtain a composite air gap response sequence.

[0073] In this embodiment, the modified low-frequency response sequence and the wave frequency response are reconstructed to obtain the composite air gap response sequence, which can be done in the following way:

[0074] The modified low-frequency response sequence and the wave frequency response are linearly superimposed in the time domain to obtain the initial reconstructed sequence;

[0075] The initial reconstructed sequence is subjected to phase consistency testing and amplitude normalization to obtain the composite air gap response sequence.

[0076] In specific implementation, firstly, the corrected low-frequency response sequence and the wave frequency response are aligned with the same time step. The amplitudes at each corresponding time point are added using array addition, and the resulting sequence is used as the initial reconstructed sequence. Then, the cross-correlation coefficient between the initial reconstructed sequence and the wave frequency response on the main frequency components is calculated. The phase consistency is then evaluated by finding the time shift corresponding to the maximum cross-correlation coefficient. For example, when the time shift exceeds 0.1 seconds, the initial reconstructed sequence is corrected accordingly. The time-shifted corrected sequence is then used as the phase-aligned sequence. The phase-aligned sequence is then normalized to obtain the composite air gap response sequence. The arithmetic mean and standard deviation of the phase-aligned sequence can be calculated using Z-score standardization. The mean and standard deviation of each data point in the sequence are then subtracted from the mean and divided by the standard deviation. The standardized sequence is then used as the composite air gap response sequence.

[0077] It should be noted that the initial reconstructed sequence in this application refers to the preliminary composite signal obtained by simple linear superposition without phase and amplitude optimization, including the corrected low-frequency and wave frequencies; phase consistency verification refers to the quality control process that verifies and ensures the correct temporal relationship between the frequency components in the reconstructed sequence through signal processing technology, which can eliminate phase deviations that may be introduced by the signal processing stage, thereby ensuring the accuracy of the time correspondence of each component in the reconstructed sequence; amplitude normalization processing refers to the standardization operation of adjusting the signal to a uniform dimension range, used to eliminate possible magnitude differences between different signal sources; the composite air gap response sequence refers to the final time-domain sequence that has undergone phase alignment and amplitude normalization processing and integrates the corrected low-frequency response and wave frequency response, which can provide a high-quality and reliable input data foundation for subsequent non-Gaussian extremum statistics.

[0078] In step S4, non-Gaussian extreme value statistics are performed based on the composite air gap response sequence to obtain the air gap extreme value prediction result of the target water area.

[0079] Preferably, in this embodiment, referenceFigure 3 As shown, this figure is an exemplary flowchart for determining the air gap extreme value prediction result of the target water area according to the information provided in this application. In this embodiment, the air gap extreme value prediction result of the target water area is obtained by performing non-Gaussian extreme value statistics based on the composite air gap response sequence, which can be achieved by the following steps:

[0080] First, in step S41, a higher-order statistical analysis is performed on the composite air gap response sequence to obtain the skewness index and the kurtosis index.

[0081] Secondly, in step S42, a polynomial transformation model based on non-Gaussian properties is constructed based on the skewness index and the kurtosis index.

[0082] Then, in step S43, Monte Carlo simulation is applied to the polynomial transformation model to generate a synthetic sequence, and then peak over-threshold sampling is performed on the synthetic sequence to obtain extreme value samples.

[0083] Finally, in step S44, the extreme value samples are fitted using a generalized Pareto distribution to obtain an extreme value distribution function, and then feature values ​​corresponding to a preset return period are extracted from the extreme value distribution function to obtain the air gap extreme value prediction result of the target water area.

[0084] In practical implementation, firstly, the ratio of the third central moment to the cube of the standard deviation of the sequence can be calculated using the skewness function, and this ratio is used as a skewness index. Then, the ratio of the fourth central moment to the fourth power of the standard deviation of the sequence can be calculated using the kurtosis function, and this ratio is used as a kurtosis index. Secondly, the standard Gaussian random variable can be used as the basic variable through a third-order Hermite polynomial expansion. The coefficients of the cubic term are determined using the skewness index, and the coefficients of the quartic term are determined using the kurtosis index. The third-order Hermite polynomial containing the basic variable, cubic coefficients, and quartic coefficients is used as a polynomial transformation model based on non-Gaussian properties. Then, Monte Carlo simulation is applied to the polynomial transformation model to generate a synthetic sequence. For example, 10,000 standard normally distributed random numbers are generated as input, and the polynomial transformation model is then applied. Converting the sequence to a non-Gaussian sequence, the resulting sequence containing 10,000 data points is used as a synthetic sequence. Based on practical air gap forecasting experience, a threshold is set at the 95th quantile of the synthetic sequence. All peak data exceeding this threshold are extracted, and these peak data are then used as extreme value samples. Finally, the extreme value samples are fitted using a generalized Pareto distribution to obtain an extreme value distribution function. Feature values ​​corresponding to a preset return period are then extracted from this extreme value distribution function to obtain the air gap extreme value prediction result for the target water area. Specifically, the scale and shape parameters of the generalized Pareto distribution can be determined using maximum likelihood estimation. The distribution function containing both scale and shape parameters is used as the extreme value distribution function. The quantiles of the extreme value distribution function are then calculated using an inverse function, and these quantile values ​​are used as the air gap extreme value prediction result for the target water area.

[0085] It should be noted that, in this embodiment, the skewness index is a third-order statistic characterizing the asymmetry of the probability distribution, used to quantify the degree to which the composite air gap response sequence deviates from the symmetrical distribution; the kurtosis index is a fourth-order statistic characterizing the steepness of the probability distribution, used to measure the thickness of the tail of the composite air gap response sequence distribution curve; the polynomial transformation model is a mathematical model that transforms a Gaussian random process into a non-Gaussian random process through polynomial expansion, used to describe the non-Gaussian nature of the composite air gap response sequence; and the extreme value distribution function is a probability distribution function describing the extreme value statistical characteristics of a random process, which can provide a quantitative basis for predicting air gap extreme values ​​under different return periods.

[0086] In step S5, a test sequence of the air gap of the semi-submersible platform changing over time is obtained through wave pool tests, and the data-driven model is iteratively optimized based on the difference between the test sequence and the predicted extreme value of the air gap.

[0087] In this embodiment, the test sequence for obtaining the time-varying air gap of the semi-submersible platform through wave pool testing can be specifically adopted in the following manner:

[0088] Construct a physical model of the semi-submersible platform;

[0089] A wave field is generated based on the wave parameters of the target water area;

[0090] The physical model is placed in the wave field, and the wave surface rise around the semi-submersible platform is measured by a wave height sensor to obtain an experimental sequence of the air gap of the semi-submersible platform changing over time.

[0091] In practical implementation, firstly, a physical model of the semi-submersible platform is constructed. This involves creating a platform model using fiberglass at a 1:50 geometric scale, ensuring that the model's mass distribution, center of gravity, and moment of inertia satisfy the Froude similarity criterion with the actual vessel. This platform entity, constructed according to the similarity criterion, serves as the physical model of the semi-submersible platform. Next, a wave field is generated based on the wave parameters of the target water area. Specifically, in a wave pool measuring 50 meters long, 30 meters wide, and 5 meters deep, a hydraulic servo wave generator system is used to input the scaling parameters corresponding to the meaningful wave height of 2.5 meters and the peak period of 8.5 seconds from the target water area to generate irregular waves. The generated waves are then processed in the wave pool. The surface motion, which simulates the wave characteristics of actual water, is used as the wave field. Finally, the physical model is placed in the wave field, and the wave rise around the semi-submersible platform is measured by wave height sensors. This yields a test sequence of the air gap of the semi-submersible platform changing over time. For example, the platform model can be moored in the center of the pool by elastic cables, and eight capacitive wave height sensors can be arranged around the four pillars of the platform. The wave time history can be recorded synchronously at a sampling frequency of 50Hz. The vertical distance between the wave surface and the bottom of the platform deck is calculated through data processing. All vertical distance measurements arranged in chronological order are then used as a test sequence of the air gap of the semi-submersible platform changing over time.

[0092] It should be noted that, in this embodiment, the physical model refers to a scaled-down platform entity made according to similarity theory that can accurately reflect the main physical characteristics of the actual ship. It can reproduce the motion and response characteristics of the real semi-submersible platform in waves in a controlled test environment. The wave field refers to a controllable water surface motion environment generated in a pool by wave-generating equipment that can simulate the actual wave conditions of the target water area. It is the physical experimental basis for verifying numerical simulation results and obtaining real air gap response data. The wave height sensor is a contact or non-contact measuring device used to measure the height of water surface fluctuations. It is used to capture the instantaneous changes of the wave surface near the semi-submersible platform and provide raw data for air gap calculation. The test sequence refers to a discrete data sequence of air gap values ​​at specific locations on the platform that change over time, obtained directly through physical model tests. It is the benchmark true value for verifying and correcting numerical simulation results.

[0093] In this embodiment, the iterative optimization of the data-driven model based on the difference between the experimental sequence and the air gap extreme value prediction results can be carried out in the following manner:

[0094] The response residual sequence is determined using the experimental sequence and the air gap extreme value prediction results;

[0095] A loss function is constructed based on the response residual sequence, and the parameter gradient of the data-driven model is calculated using the gradient descent algorithm.

[0096] The internal weight parameters of the data-driven model are iteratively updated according to the gradient of the parameters until the loss function converges, thus completing the iterative optimization of the data-driven model.

[0097] In practice, firstly, the experimental sequence can be used as the baseline truth. The air gap extreme value prediction result is extended into a time series of the same length as the experimental sequence. Then, the difference between the two sequences is calculated point by point using array subtraction. The array of all differences arranged in chronological order is then used as the response residual sequence. Next, the mean squared error formula is used to sum the squared differences of all data points in the response residual sequence and take the average value. The result is then used as the loss function. The stochastic gradient descent algorithm is then used to calculate the partial derivatives of the loss function with respect to all weight parameters in the data-driven model using a learning rate of 0.001. These partial derivative values ​​are then used as the parameter gradients. Finally, the internal weight parameters of the data-driven model are iteratively updated according to the parameter gradients. That is, the current weight parameters are subtracted from the product of the learning rate and the parameter gradients to obtain new weight parameter values. This process is repeated until the curve of the change in the loss function value converges to smoothness in 10 consecutive iterations. At this point, the above adjustment process can be used as the iterative optimization of the data-driven model.

[0098] It should be noted that, in this embodiment, the response residual sequence refers to the sequence of differences between the experimental measured values ​​and the model predicted values, used to quantify the degree of deviation between the current predictive ability of the data-driven model and the actual physical response; the loss function is a mathematical function used to evaluate the predictive performance of the model, and in this application, it specifically refers to the mean squared error function constructed based on the response residuals, and the magnitude of the loss function represents the prediction accuracy of the data-driven model; the parameter gradient refers to the rate of change of the loss function relative to the model's weight parameters, which can provide accurate adjustment direction and magnitude for the parameter updates of the data-driven model; the internal weight parameters refer to the connection strength values ​​between neurons in the data-driven model, and by iteratively optimizing these parameters, the model can gradually approach the actual physical laws; convergence refers to the state in which the loss function value tends to stabilize and no longer decreases significantly during the optimization process, indicating that the data-driven model has reached its optimal performance.

[0099] In summary, the technical solution adopted in this application can achieve data-driven precise correction of low-frequency response in numerical simulation, thereby improving the accuracy of air gap extreme value prediction for semi-submersible platforms.

[0100] Example 2: This application provides a semi-submersible platform air gap extreme value prediction system based on numerical simulation, referencing... Figure 4 As shown in the figure, this is a module structure diagram of a semi-submersible platform air gap extreme value prediction system based on numerical simulation provided in this application. The prediction system includes:

[0101] The numerical simulation module 100 is used to obtain the air gap response sequence of the semi-submersible platform in the target water area through numerical simulation. The air gap response sequence includes wave frequency response and low frequency response.

[0102] The low-frequency correction module 200 is used to extract the low-frequency response from the air gap response sequence, and perform nonlinear correction on the low-frequency response based on a data-driven model to obtain a corrected low-frequency response sequence.

[0103] The response reconstruction module 300 is used to reconstruct the corrected low-frequency response sequence and the wave frequency response to obtain a composite air gap response sequence.

[0104] The extreme value prediction module 400 is used to perform non-Gaussian extreme value statistics based on the composite air gap response sequence to obtain the air gap extreme value prediction result of the target water area.

[0105] The model optimization module 500 is used to obtain a test sequence of the air gap of a semi-submersible platform changing over time through wave pool tests, and to iteratively optimize the data-driven model based on the difference between the test sequence and the predicted extreme values ​​of the air gap.

[0106] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compactdisc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0108] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A method for predicting air gap extreme values ​​of a semi-submersible platform based on numerical simulation, characterized in that, The forecasting method includes the following steps: The air gap response sequence of the semi-submersible platform in the target water area was obtained by numerical simulation. The air gap response sequence includes wave frequency response and low frequency response. The low-frequency response is extracted from the air gap response sequence, and the low-frequency response is nonlinearly corrected based on a data-driven model to obtain the corrected low-frequency response sequence. The modified low-frequency response sequence and the wave frequency response are reconstructed to obtain the composite air gap response sequence. Based on the composite air gap response sequence, non-Gaussian extreme value statistics are performed to obtain the air gap extreme value prediction results for the target water area. A test sequence of air gap variation over time for a semi-submersible platform was obtained through wave pool tests. The data-driven model was iteratively optimized based on the difference between the test sequence and the predicted extreme values ​​of the air gap.

2. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, The air gap response sequence of a semi-submersible platform in the target water area obtained through numerical simulation specifically includes: The transfer functions of motion and wave load of the semi-submersible platform are constructed based on the structural parameters of the semi-submersible platform and the wave parameters of the target water area. The wave frequency motion response and wave load are determined by the transfer function. The air gap response sequence is obtained by performing coupled analysis on the wave frequency motion response and wave load.

3. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, The low-frequency response is extracted from the air gap response sequence based on a high-pass filter.

4. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, The data-driven model is a machine learning model based on historical data, used to learn and correct the inherent low-frequency response bias in the numerical simulation through the modal features of the low-frequency response.

5. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, Based on a data-driven model, the low-frequency response is nonlinearly corrected to obtain the corrected low-frequency response sequence, which specifically includes: Modal decomposition of the low-frequency response yields multiple intrinsic mode function components; Construct a training set based on all intrinsic mode function components; All intrinsic mode function components are input into the data-driven model trained on the training set for signal reconstruction, resulting in the corrected low-frequency response sequence.

6. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, Reconstructing the modified low-frequency response sequence and the wave frequency response to obtain the composite air gap response sequence specifically includes: The modified low-frequency response sequence and the wave frequency response are linearly superimposed in the time domain to obtain the initial reconstructed sequence; The initial reconstructed sequence is subjected to phase consistency testing and amplitude normalization to obtain the composite air gap response sequence.

7. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, Based on the composite air gap response sequence, non-Gaussian extreme value statistics are performed to obtain the air gap extreme value prediction results for the target water area, specifically including: Higher-order statistical analysis was performed on the composite air gap response sequence to obtain skewness and kurtosis indices. A polynomial transformation model based on non-Gaussian properties is constructed based on the skewness index and the kurtosis index. Monte Carlo simulation is applied to the polynomial transformation model to generate a synthetic sequence, and then peak over-threshold sampling is performed on the synthetic sequence to obtain extreme value samples; The extreme value samples are fitted using a generalized Pareto distribution to obtain an extreme value distribution function. Then, feature values ​​corresponding to a preset return period are extracted from the extreme value distribution function to obtain the air gap extreme value prediction result for the target water area.

8. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, The test sequence for obtaining the time-varying air gap of a semi-submersible platform through wave tank tests specifically includes: Construct a physical model of the semi-submersible platform; A wave field is generated based on the wave parameters of the target water area; The physical model is placed in the wave field, and the wave surface rise around the semi-submersible platform is measured by a wave height sensor to obtain an experimental sequence of the air gap of the semi-submersible platform changing over time.

9. The method for predicting the extreme values ​​of air gap in a semi-submersible platform based on numerical simulation as described in claim 1, characterized in that, Iterative optimization of the data-driven model based on the difference between the experimental sequence and the air gap extreme value prediction results specifically includes: The response residual sequence is determined using the experimental sequence and the air gap extreme value prediction results; A loss function is constructed based on the response residual sequence, and the parameter gradient of the data-driven model is calculated using the gradient descent algorithm. The internal weight parameters of the data-driven model are iteratively updated according to the gradient of the parameters until the loss function converges, thus completing the iterative optimization of the data-driven model.

10. A numerical simulation-based air gap extreme value prediction system for semi-submersible platforms, used to execute the numerical simulation-based air gap extreme value prediction method for semi-submersible platforms as described in any one of claims 1 to 9, characterized in that, The forecasting system includes: The numerical simulation module is used to obtain the air gap response sequence of the semi-submersible platform in the target water area through numerical simulation. The air gap response sequence includes wave frequency response and low frequency response. The low-frequency correction module is used to extract the low-frequency response from the air gap response sequence, and perform nonlinear correction on the low-frequency response based on a data-driven model to obtain a corrected low-frequency response sequence. The response reconstruction module is used to reconstruct the corrected low-frequency response sequence and the wave frequency response to obtain a composite air gap response sequence. The extreme value prediction module is used to perform non-Gaussian extreme value statistics based on the composite air gap response sequence to obtain the air gap extreme value prediction result of the target water area; The model optimization module is used to obtain a test sequence of the air gap of a semi-submersible platform changing over time through wave pool tests, and to iteratively optimize the data-driven model based on the difference between the test sequence and the predicted extreme values ​​of the air gap.

Images