Simulation method for instability research of seabed data cabin structure under action of wave current and earthquake coupling load

Abstract

The invention discloses a simulation method for instability research of a seabed data cabin structure under the action of wave current and earthquake coupling load, and belongs to the field of seabed data cabin instability research, and the method comprises the steps: constructing a three-dimensional structure model, obtaining wave current parameters and earthquake waveform data, and generating a coupling load data set; performing computational fluid mechanics simulation based on the data set, and integrating seismic waveforms to obtain nonlinear mechanical effect distribution; when the flow velocity distribution exceeds a threshold value, an iterative optimization algorithm is adopted to adjust a wave height parameter and a seismic intensity coefficient, and a stable wave flow coupling field is obtained; extracting a structural acting force vector from a stable result, predicting displacement and stress response by using a neural network model, and judging whether the displacement and stress response meet an instability risk threshold value or not; if a threshold value is met, fusing multi-sensor signals to generate a real-time mechanical response data sequence; performing frequency domain analysis on the data sequence to determine a vibration mode; and extracting an instability index from the vibration mode, and determining an instability type and an occurrence probability by adopting a classification algorithm.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of research on the instability of submarine data capsules, and particularly relates to a simulation method for researching the instability of a submarine data capsule structure under the action of wave-current and earthquake coupling loads. BACKGROUND

[0002] Research in the field of ocean engineering is of key significance to the protection of marine resource development and infrastructure safety. With the increasing global demand for deep-sea resources, submarine data capsules, as the core equipment for storing and processing marine data, their stability and reliability in extreme marine environments directly affect the success or failure of ocean engineering projects. Submarine data capsules need to withstand the combined action of various loads such as ocean currents, waves and earthquakes in complex and changing marine environments, ensuring that the structure does not overturn or be damaged. However, existing research methods often face technical limitations when simulating these complex environmental conditions, and cannot fully reveal the mechanical response and instability mechanism of submarine data capsules in real environments, which poses a great challenge to structural design and safety assessment.

[0003] Current research methods mainly rely on single or static load condition analysis when simulating the stress behavior of submarine data capsules, such as considering only the individual action of ocean currents or waves. Although this method simplifies the experimental design, it ignores the mutual influence of multiple loads acting simultaneously, resulting in a large deviation between the simulation results and the complexity of actual marine environments. Especially in the scenario of wave-current coupling and earthquake load acting together, existing methods are difficult to accurately capture the response characteristics of the structure in dynamic environments. This limitation makes researchers often underestimate the complexity and unpredictability of environmental loads when predicting the instability risk of submarine data capsules, thereby affecting the reliability of the design.

[0004] Therefore, how to accurately reproduce the wave-current coupling field in the laboratory and apply controllable seismic loads in dynamic environments while obtaining real-time mechanical response data of the structure becomes a key problem in the field of ocean engineering research on the stability of submarine data capsules. SUMMARY

[0005] To solve the above technical problems, the present application provides a simulation method for researching the instability of a submarine data capsule structure under the action of wave-current and earthquake coupling loads, comprising:

[0006] A three-dimensional structural model of the submarine data capsule is constructed by a finite element analysis method, and initial wave-current parameters and seismic waveform data are obtained from a pre-set marine environment database to generate a coupling load data set;

[0007] Computational fluid dynamics simulation is performed according to the coupling load data set to obtain a dynamic wave-current coupling field, and seismic waveform data is integrated during the simulation process to form a nonlinear mechanical effect distribution;

[0008] When the flow velocity distribution in the coupling load dataset exceeds a preset threshold value, an iterative optimization algorithm is used to adjust the wave height parameter and the seismic intensity coefficient to obtain a stable wave flow coupling field simulation result;

[0009] A structure action force vector is extracted from the stable wave flow coupling field simulation result, a neural network model is used to predict the displacement and stress response of the seabed data capsule in a dynamic environment, and it is determined whether a loss of stability risk threshold value is met;

[0010] If the loss of stability risk threshold value is met, a multi-sensor simulation signal is fused to generate a real-time mechanical response data sequence;

[0011] Frequency domain analysis is performed according to the real-time mechanical response data sequence, spectral features are extracted, and a structure vibration mode is determined;

[0012] Key instability indicators are extracted from the structure vibration mode, a classification algorithm is used for processing, and a loss of stability type and occurrence probability are obtained;

[0013] According to the classification result, the load parameters in the marine environment database are updated to obtain an optimized coupling load dataset.

[0014] Optionally, the three-dimensional structure model comprises:

[0015] A finite element method is used to establish a seabed data capsule geometric model, grid division and material properties are set;

[0016] Wave flow parameters and seismic wave form data are obtained from the database, and standardized parameter sets are obtained through data cleaning;

[0017] Fluid dynamics simulation and frequency spectrum analysis are performed on the standardized parameter sets to obtain flow velocity distribution and vibration frequency data respectively, and the coupling load dataset is generated.

[0018] Optionally, the computational fluid dynamics simulation comprises:

[0019] Time sequence features of the seismic wave form data are extracted to obtain vibration signal input;

[0020] The vibration signal is superimposed on the wave flow coupling field to generate superimposed flow field data;

[0021] Flow direction changes and nonlinear mechanical effect distributions are determined according to the superimposed flow field data.

[0022] Optionally, the iterative optimization algorithm comprises:

[0023] When the flow velocity distribution exceeds the threshold value, flow velocity abnormal points are extracted and a flow velocity abnormal data set is generated;

[0024] Based on the flow rate abnormal data set, an optimization algorithm is used to adjust the wave height parameter and the seismic intensity coefficient;

[0025] According to the adjusted parameters, the stable wave-current coupling field simulation result is generated.

[0026] Optionally, the predicted displacement and stress response includes:

[0027] The force vector acting on the structure is decomposed to obtain force vector components;

[0028] The neural network model is trained by taking the force vector components as input data;

[0029] The neural network model is used to calculate the displacement response prediction value and the stress response prediction value.

[0030] Optionally, the fusion multi-sensor simulation signal includes:

[0031] The vibration frequency and stress change signal collected by the sensor are preprocessed to generate a denoising signal sequence;

[0032] The denoising signal sequence is fused by using a weighted average method to form a unified multi-dimensional fusion data sequence.

[0033] Optionally, the extraction of the key instability index includes:

[0034] The vibration mode signal is subjected to Fourier transform to extract frequency components and amplitude characteristics;

[0035] The principal component analysis method is used to extract key indicators to form an index set;

[0036] The index set is classified by using a support vector machine classifier to obtain an instability type and probability.

[0037] Optionally, the updating of the load parameter includes:

[0038] The initial load parameter is determined according to the classification result;

[0039] The load parameter is subjected to cluster analysis to generate an optimization data set;

[0040] The optimization data set is simulated by iteration to obtain a final optimized coupling load data set.

[0041] In another aspect, the present application also provides an electronic device, which includes a memory, a processor, and a computing program stored in the memory and executable on the processor, and the processor implements the method when executing the computing program.

[0042] In another aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method.

[0043] Compared with the prior art, the present application has the following advantages and technical effects:

[0044] The application discloses a method for structural stability analysis and instability risk prediction of a seabed data capsule under a complex marine environment, and solves the problem of accurate prediction and optimization of structural response under the combined action of dynamic wave flow coupling and seismic load. The application constructs a three-dimensional model of the seabed data capsule through finite element analysis, generates a coupling load data set in combination with wave flow parameters and seismic waveform data of a marine environment database, and simulates a dynamic wave flow coupling field by using computational fluid dynamics, and integrates seismic data to capture nonlinear mechanical effects. When the flow rate exceeds a threshold value, the application adjusts the wave height and seismic intensity parameters by using an iterative optimization algorithm to ensure the stability of the simulation results. Based on the stable simulation results, the application extracts a structural force vector, predicts displacement and stress response by using a neural network, and judges the instability risk. If the risk exceeds a standard, the application integrates multi-sensor signals by using a data fusion technology to generate a real-time mechanical response sequence, and performs frequency domain analysis to extract a vibration mode and a key instability index. A support vector machine classifier further judges the instability type and probability, and optimizes the load parameters to improve the simulation accuracy in the subsequent simulation. The application significantly improves the stability evaluation and risk prevention and control capability of the seabed data capsule under extreme environments, and provides reliable technical support for deep sea engineering design. BRIEF DESCRIPTION OF DRAWINGS

[0045] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application and serve to explain the principle of the present application and its advantages to those skilled in the art, and are provided to describe the illustrative embodiments of the present application together with the description. In the drawings:

[0046] Figure 1 The method flowchart of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0047] It should be noted that the embodiments and the features in the embodiments in the present application can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a group of computer executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown herein.

[0049] Embodiment one

[0050] As Figure 1As shown, this embodiment provides a simulation method for studying the instability of a seabed data cabin structure under coupled wave-current and seismic loads, including:

[0051] A three-dimensional structural model of the seabed data cabin was constructed using the finite element analysis method, and initial wave and current parameters and seismic waveform data were obtained from a pre-set marine environment database to generate a coupled load dataset.

[0052] Computational fluid dynamics simulations are performed based on the coupled load dataset to obtain a dynamic wave-current coupled field. Seismic waveform data is integrated during the simulation process to form a nonlinear mechanical effect distribution.

[0053] When the velocity distribution in the coupled load dataset exceeds a preset threshold, an iterative optimization algorithm is used to adjust the wave height parameter and the seismic intensity coefficient to obtain stable wave-current coupled field simulation results.

[0054] The structural force vector is extracted from the simulation results of the stable wave-current coupling field. The displacement and stress response of the seabed data cabin in a dynamic environment are predicted using a neural network model, and it is determined whether the instability risk threshold is met.

[0055] If the instability risk threshold is met, multi-sensor simulation signals are fused to generate a real-time mechanical response data sequence.

[0056] Frequency domain analysis is performed on the real-time mechanical response data sequence to extract spectral features and determine the structural vibration mode;

[0057] Key instability indicators are extracted from the vibration modes of the structure, and a classification algorithm is used to process them to obtain the instability type and probability of occurrence.

[0058] The load parameters in the marine environment database are updated based on the classification results to obtain the optimized coupled load dataset.

[0059] S101. Construct a three-dimensional structural model of the seabed data cabin using the finite element analysis method, obtain initial wave and current parameters and seismic waveform data from a pre-set marine environment database, and obtain a coupled load dataset containing velocity distribution and vibration frequency.

[0060] The embodiment first constructs a large and high-stability water tank. The water tank is made of high-strength and corrosion-resistant special steel to effectively resist seawater erosion and various loads in long-term complex marine environment simulation. The size of the water tank is carefully designed according to actual research needs. The length, width and depth of the water tank should be sufficient to accommodate the scaled model of the seabed data capsule and various equipment required for simulation, while ensuring that the propagation of wave flow and the effect of seismic waves are not disturbed by boundary conditions. For example, the length of the water tank can be designed to be 40-80 meters, the width can be 8-16 meters, and the depth can be 4-8 meters. The bottom of the water tank is equipped with a special reinforcement structure that can withstand the strong seismic load transmitted by the shaking table. The bottom is provided with a drainage and dredging system to facilitate cleaning and maintenance of the water tank before and after the test. The specific implementation of the water tank inspection and cleaning requirements are as follows: use a high-pressure water gun and special cleaning agent to thoroughly flush the inside of the water tank to remove residual dirt, rust and debris. Check the welded parts and sealing strips of the water tank to ensure that there are no cracks or leaks. If a small crack is found in the water tank, it can be repaired by welding and polishing. For aging of the sealing strip, replace the new seawater corrosion-resistant strip in time. Check the reinforcement structure and drainage and dredging system at the bottom of the water tank to ensure that the drainage pipeline is unobstructed and the dredging equipment is operating normally.

[0061] An advanced multi-directional wave maker is installed at one end of the water tank. The wave maker uses the latest innovative structure of push plate combined with swing plate, which can generate various types of waves, including regular waves, irregular waves, extreme waves, etc. Through a high-precision computer control system, the wave height, wavelength, period and propagation direction of the wave can be accurately adjusted to simulate the characteristics of waves under different sea conditions. For example, the wave maker can generate waves with a wave height range of 0.05-1.5 meters, a wavelength range of 1-15 meters, and a period range of 0.5-8 seconds, and the wave propagation direction can be flexibly adjusted within 0-360 degrees.

[0062] To ensure the stability and uniformity of the waves, the wave maker is equipped with high-sensitivity wave sensors that monitor wave parameters in real time and make precise adjustments to the movement of the wave maker through a feedback control system. The specific calibration steps are as follows: fill the water tank with water to the designed water level, start the wave maker to generate regular waves, set the wave height to 0.3 meters, the wavelength to 4 meters, and the period to 1.5 seconds, and run for 10 minutes. Use a wave height gauge to measure the wave height at different positions (front, middle, back, left, middle, right) in the water tank. If the deviation of the wave height at each point from the set value exceeds ±3%, adjust the movement parameters of the push plate and swing plate through the wave maker control system until the wave height uniformity meets the standard. Switch to irregular wave mode, set the JONSWAP spectrum parameters, and run for 20 minutes. Observe the wave shape to ensure that there are no obvious abnormal fluctuations in the wave field.

[0063] A set of high-efficiency water flow driving devices is installed in the water tank, which is composed of multiple high-power, low-noise water pumps and a carefully designed flow guide system. By precisely adjusting the speed and flow rate of the water pumps, a stable and controllable water flow can be generated to simulate different flow rates and directions of ocean currents. The flow guide system adopts a special streamlined design to ensure uniform distribution of water flow in the tank, effectively avoiding the generation of water vortex and turbulence. For example, the water flow driving device can generate a stable water flow with a flow rate ranging from 0.1 to 2.5 meters per second, and the water flow direction can be freely adjusted within 0-360 degrees according to the test requirements. By precisely controlling the operating parameters of the water pumps, various complex ocean current conditions such as tidal currents and coastal currents can be simulated.

[0064] The specific commissioning method of the water flow driving device is as follows: start the water pump, gradually increase the flow rate from low speed to 1 meter per second, and observe the water flow pattern to ensure that there is no obvious vortex and turbulence. Multiple flow rate meters are arranged in the water tank to monitor the water flow speed and direction in real time. If the flow rate deviation exceeds ±0.1 meters per second or the flow direction deviation exceeds ±5°, adjust the water pump speed, flow guide plate angle, or add auxiliary flow guide devices for calibration until the water flow is uniform and stable.

[0065] An ultra-high performance vibration table is installed at the center of the bottom of the water tank. The vibration table adopts advanced electromagnetic drive technology, with excellent control accuracy and strong load capacity. The vibration table can accurately simulate various types of seismic waves, including seismic waves of different magnitudes, frequencies, and durations. Through advanced control systems, the vibration table can accurately control the motion according to pre-set seismic wave parameters such as sine wave, triangle wave, and actual seismic record wave, to realize real simulation of different seismic conditions. For example, the maximum acceleration of the vibration table can reach more than 6g, the frequency range is 0.1-120Hz, and it can carry a seabed data capsule scale model weighing several tons. The vibration table is also equipped with high-precision acceleration sensors and displacement sensors to monitor the motion state of the vibration table in real time, ensuring the accuracy and reliability of the seismic simulation.

[0066] A special installation area for seabed data capsule models is set up in the water tank, which is equipped with various types of model fixing and connecting devices to adapt to different structural forms and fixing methods of seabed data capsule models. For seabed data capsule models fixed with pile foundations, special pile holes and pile fixing clamps are provided to ensure that the pile foundation model can be firmly fixed to the bottom of the water tank and accurately simulate the interaction between pile and soil. For floating seabed data capsule models fixed with anchoring systems, corresponding anchor points and mooring cable connecting devices are installed to accurately simulate the stress state of the anchoring system under the action of waves and currents. At the same time, the model fixing and connecting devices have high adjustability, which can easily adjust the installation angle and position of the model to meet the needs of different test conditions.

[0067] A high-precision scale model of the seabed data capsule is made, and the model is designed and manufactured in strict accordance with the similarity criteria to ensure that its mechanical properties and actual structure have high similarity. The model includes the data capsule body, support structure, connecting components and other parts, and each part is made of a material similar in mechanical properties to the actual material, such as aluminum alloy or high-strength engineering plastic to simulate metal structures, and flexible materials to simulate the elastic deformation of connecting components.

[0068] By the finite element analysis method, a three-dimensional structure model of the seabed data capsule is generated by using a structure modeling software to determine the mesh division and material properties, and an initial three-dimensional structure model is obtained. From a pre-set ocean environment database, wave flow parameters and seismic waveform data are obtained, and missing values and abnormal values are processed by using data cleaning techniques to obtain a standardized environment parameter set. For the standardized environment parameter set, fluid dynamics simulation is used to calculate the flow velocity distribution, and a flow field data set containing time series is generated. By using the seismic waveform data, frequency spectrum analysis method is used to extract vibration frequency characteristics, and a frequency distribution data set is obtained. The flow field data set and the frequency distribution data set are combined, and the fluid-structure coupling effect is calculated by using a coupling analysis algorithm to generate a coupled load data set. If the load value in the coupled load data set exceeds the pre-set threshold value, the mesh parameters of the three-dimensional structure model are adjusted by using an iterative optimization method, and the coupled load data set is recalculated to obtain an optimized coupled load data set. According to the optimized coupled load data set, the structural response of the seabed data capsule is analyzed by using a finite element simulation software to obtain stress distribution and deformation data.

[0069] For example, a three-dimensional structure model of the seabed data capsule is constructed by the finite element analysis method. First, a CAD software (such as ANSYS SpaceClaim) is used to generate a data capsule geometric model, assuming that the capsule body is a cylindrical shell with a diameter of 2.5 meters and a length of 10 meters, a wall thickness of 0.02 meters, and a material of high-strength steel (Young's modulus of 210 GPa, Poisson's ratio of 0.3). SolidWorks is used to generate the initial mesh, and the mesh size is set to 0.05 meters to ensure that the number of nodes is about 500,000 to balance the precision and calculation efficiency. After mesh division, the geometric model is imported into ANSYS Mechanical using tetrahedral elements (Solid187) for finite element discretization. The initial wave flow parameters are obtained from a pre-set ocean environment database, and a typical working condition is selected: wave height of 2 meters, period of 8 seconds, flow velocity of 0.5 meters / second, and direction along the axis of the capsule body. The wave flow load is calculated using the Morison equation, which is F = ρC_dAu|u| + ρC_mVü, where ρ is the density of seawater 1025 kg / m 3 , C_d is the drag coefficient 1.2, C_m is the inertia coefficient 2.0, u is the flow velocity, and ü is the acceleration. The distributed load is calculated to be about 10 kN / m 2The seismic waveform data is extracted from the database, and the El Centro earthquake wave with a peak acceleration of 0.3g and a duration of 30 seconds and a spectrum range of 0.1-10Hz is selected. The seismic wave spectrum is analyzed by using the fast Fourier transform (FFT) algorithm, and the main frequency of 2.5Hz is extracted as the vibration excitation input. The coupling load data set is calculated by coupling ANSYS Fluent and Mechanical modules, the k-ε turbulence model is used for fluid dynamics, the inlet flow velocity is set to 0.5m / s, the wall boundary is set to no-slip condition, and the iteration is 1000 times to ensure convergence (residual <10 -5 ) The structure response is solved by Newmark-β method, the time step is 0.01s, the vibration frequency distribution is calculated, and it is found that the main vibration mode frequency is 2.7Hz, which is close to the main frequency of the seismic wave, indicating that there is a resonance risk. The final output includes the coupling data set of flow velocity distribution (maximum 0.52m / s) and vibration frequency (2.7Hz), stored in CSV format, used for subsequent optimization design.

[0070] S102, generate a dynamic wave-flow coupling field by using computational fluid dynamics simulation according to the coupling load data set, integrate the seismic waveform data in the simulation process, and determine the nonlinear mechanical effect distribution of the flow direction change and vibration superposition in the wave-flow coupling field.

[0071] The wave-flow coupling field is simulated by using the computational fluid dynamics method to obtain the initial flow field data. The time sequence characteristics are extracted from the seismic waveform data to obtain the vibration signal input. The vibration signal input is superimposed on the wave-flow coupling field by data integration processing to generate the superimposed flow field data. The flow direction change characteristics are analyzed for the superimposed flow field data to determine the flow direction change distribution. If the flow direction change distribution exceeds the preset threshold, the nonlinear mechanical effect is calculated by using the finite element analysis method to obtain the effect distribution data. According to the effect distribution data, the k-means clustering algorithm is used to divide the effect area to obtain the area distribution result. The nonlinear mechanical effect distribution is generated by the area distribution result to determine the final distribution characteristics.

[0072] Exemplarily, based on the coupled load data set, a dynamic wave-current coupling field is generated by using a computational fluid dynamics (CFD) simulation. The Fluent software is used to solve the three-dimensional Navier-Stokes equation. The flow field grid is set to 1 million tetrahedral elements, the grid size is 0.01 m, and the time step is 0.001 s. The initial conditions are a flow velocity of 1.5 m / s, a wave height of 0.5 m, and a period of 2 s. The k-ε turbulence model is used, and the boundary conditions are fixed inlet velocity, outlet pressure boundary, and atmospheric pressure of 101325 Pa. In the simulation, the seismic waveform data is integrated. It is assumed that the seismic wave is a sine wave with a frequency of 5 Hz and an amplitude of 0.02 m. Based on the time domain superposition method, the seismic vibration displacement is directly applied to the flow field bottom boundary. The finite volume method is used for discretization and solving. The fluid velocity and pressure distribution at each time step are calculated to generate the wave-current coupling field. When the flow direction changes, the X-Y plane velocity vector in the flow field is extracted, and the flow direction angle change rate is calculated. The maximum change rate is 15° / s, indicating that the wave-current interaction significantly enhances the local turbulence. After the vibration superposition, the nonlinear momentum equation is used to analyze the mechanical effect, and the Reynolds stress distribution is calculated. It is found that the stress peak value in the near-wall region is 200 Pa, indicating that the seismic vibration induces significant nonlinear effects. During the analysis, the velocity gradient and vorticity field are extracted by using the post-processing software. The maximum vorticity is 50 s -1 , indicating that the vibration-induced vortex structure is enhanced. Logically, the CFD simulation provides the basic flow field data, the seismic waveform superposition introduces the vibration effect, the flow direction change and the nonlinear mechanical effect are quantified by numerical analysis, and the rigorous derivation from the data set to the final distribution is ensured. If the data set is missing, it can be supplemented by synthesizing the waveform (such as a sine wave) while maintaining the parameter consistency.

[0073] S103, if the flow velocity distribution in the coupled load data set exceeds the preset threshold value, the wave height parameter and the seismic intensity coefficient are adjusted by using an iterative optimization algorithm to obtain stable wave-current coupling field simulation results, wherein the iterative optimization algorithm processes the parameter adjustment accuracy to match the real sea conditions.

[0074] If the flow velocity distribution exceeds the preset threshold, the data processing module extracts the flow velocity abnormal points in the coupled load data set to obtain a flow velocity abnormal data set. According to the flow velocity abnormal data set, a particle swarm optimization algorithm is used to adjust the wave height parameters to generate a first wave height parameter set. If the matching degree of the first wave height parameter set with the real marine environment is lower than the preset standard, a gradient descent algorithm is used to optimize the seismic intensity coefficient to obtain a first seismic intensity coefficient set. According to the first wave height parameter set and the first seismic intensity coefficient set, a preliminary wave-current coupling field simulation result is generated. If the stability of the preliminary wave-current coupling field simulation result is lower than the preset standard, an iterative optimization algorithm is used to jointly adjust the wave height parameters and the seismic intensity coefficient to obtain a second wave height parameter set and a second seismic intensity coefficient set. According to the second wave height parameter set and the second seismic intensity coefficient set, a final wave-current coupling field simulation result is generated. The environmental matching module compares the final wave-current coupling field simulation result with the real marine environment data to determine the matching accuracy of the simulation result.

[0075] For example, for the flow velocity distribution in the coupled load data set, first, the data preprocessing module analyzes the marine flow velocity data, assuming that the data set contains flow velocity values (unit: m / s), and setting the preset threshold to 2.5 m / s. The NumPy library in Python is used to calculate the mean and standard deviation of the flow velocity.

[0076] For example, the data set contains 1000 flow velocity data points, the mean is calculated to be 2.0 m / s, and the standard deviation is calculated to be 0.3 m / s. If the proportion of data points exceeding 2.5 m / s is more than 10%, the parameter adjustment is triggered. Next, the particle swarm optimization (PSO) algorithm is used to adjust the wave height parameters and the seismic intensity coefficient, the wave height parameter is initialized to 1.5 m, the seismic intensity coefficient is initialized to 0.8, the PSO algorithm is set to have 50 particles, 100 iterations, an inertia weight of 0.7, and learning factors c1=c2=2.0. The stability of the wave-current coupling field is evaluated by the objective function, which is defined as the mean square error (MSE) of the flow velocity distribution and the real marine conditions, and the MSE is required to be less than 0.01. During the iteration process, the PSO algorithm adjusts the wave height parameter (range: 1.0-2.0 m) and the seismic intensity coefficient (range: 0.5-1.0) according to the flow velocity distribution. The fluid mechanics equation (Navier-Stokes equation) of the coupling field is calculated every iteration, and the finite element method is used for solution with a grid division precision of 0.01 m. After 50 iterations, the wave height parameter converges to 1.65 m, the seismic intensity coefficient converges to 0.75, and the MSE decreases to 0.008, satisfying the stability condition.

[0077] S104, from the stable wave-current coupling field simulation result, extract the structural action force vector, use the neural network model to predict the displacement and stress response of the seabed data capsule in the dynamic environment, and judge whether the predicted response meets the instability risk threshold.

[0078] The structural force vector is extracted from the wave-current coupled field simulation results, and the force vector is decomposed by numerical analysis method to obtain the force vector component. According to the force vector component, the input data set of the neural network model is constructed, the multilayer perception model is trained to obtain the displacement response prediction value and the stress response prediction value. Through the displacement response prediction value and the stress response prediction value, the dynamic response characteristics of the seabed data capsule are calculated, and the dynamic response distribution is obtained. Using a preset instability risk threshold, if the displacement response prediction value in the dynamic response distribution exceeds the preset displacement threshold, it is marked as high displacement risk, and the displacement risk assessment result is obtained. Using a preset instability risk threshold, if the stress response prediction value in the dynamic response distribution exceeds the preset stress threshold, it is marked as high stress risk, and the stress risk assessment result is obtained. Through the displacement risk assessment result and the stress risk assessment result, combined with logical and operation, if any risk is marked as high, it is determined that the seabed data capsule has instability risk, and the instability risk determination result is obtained. According to the instability risk determination result, the state classification of the seabed data capsule is carried out by using a classification algorithm, and the final state classification result is obtained.

[0079] For example, the structural force vector is extracted from the stable wave-current coupled field simulation results. First, the seabed environment is numerically simulated by computational fluid dynamics software (such as ANSYS Fluent), the sea current velocity is set to 1.2 m / s, the wave period is set to 8 seconds, and the wave height is set to 2 meters to generate three-dimensional flow field data. The finite volume method is used to discretize the Navier-Stokes equation, and the mesh is divided into 1 million tetrahedral elements to ensure that the convergence residual is less than 10^-5. When extracting the force vector, the surface mesh of the seabed data capsule is selected, the pressure and shear stress of each mesh element are calculated, and the total force vector is obtained, for example, Fx=5000N, Fy=2000N, Fz=3000N. Subsequently, a neural network model is constructed to predict the displacement and stress response, a BP neural network is used, which includes 3 layers: input layer (6 nodes, including force vector and environmental parameters), hidden layer (20 nodes, ReLU activation function), output layer (4 nodes, output displacement x, y, z and maximum stress). The training data set comes from 1000 simulation data, the training set and test set ratio is 8:2, the optimization algorithm uses Adam, the learning rate is 0.001, the iteration is 500 times, and the loss function mean square error converges to 0.01. The prediction results show that the data capsule has a displacement of 0.05m in the x direction, 0.02m in the y direction, 0.01m in the z direction, and a maximum stress of 150MPa. Finally, the instability risk is judged, the displacement threshold is set to 0.1m, and the stress threshold is set to 200MPa. The predicted displacement and stress do not exceed the threshold, indicating that there is no instability risk. To verify the reliability, 1000 random wave-current conditions are generated by the Monte Carlo method, the prediction is repeated, and 95% of the results meet the threshold requirements, confirming that the prediction model is robust.

[0080] S105, if the predicted response meets the instability risk threshold, merging the multi-sensor analog signals by data fusion techniques to obtain a real-time mechanical response data sequence, wherein the data fusion techniques integrate vibration frequencies and stress changes to reveal potential instability mechanisms.

[0081] If the multi-sensor collected analog signals contain vibration frequencies and stress changes, the signals are preprocessed by Kalman filtering algorithm to obtain denoised sensor data sequence. According to the denoised sensor data sequence, the vibration frequencies and stress changes are fused by weighted average method to generate a unified multi-dimensional fusion data sequence. If the vibration frequency or stress change in the fusion data sequence exceeds the preset threshold, the main features are extracted by principal component analysis algorithm to obtain a feature vector representing the instability risk. According to the feature vector, a pre-established classification model is used to judge the instability risk level to obtain the risk assessment result. If the risk assessment result shows high risk, the dynamic correlation of vibration frequency and stress change is calculated according to the time variation trend of the fusion data sequence to obtain a quantitative index of instability mechanism. According to the quantitative index, time series analysis method is used to predict the instability probability in the future period of time to obtain a prediction data sequence. According to the prediction data sequence, an updated sequence of real-time mechanical response data is generated to obtain an optimized system state description.

[0082] For example, after the predicted response meets the instability risk threshold, the system first determines whether the structure reaches the instability risk by setting a threshold, such as vibration frequency exceeding 50 Hz or stress change rate greater than 10 MPa / s, triggering the data fusion process. The analog signals collected by the sensors include vibration frequency measured by the accelerometer and stress value measured by the strain gauge, and the data is sampled at 10 ms intervals to generate a time series. The multi-sensor data is fused by Kalman filtering algorithm, assuming that the accelerometer data is x1(t) = 50sin(2πft) + ε1 and the strain gauge data is x2(t) = 12cos(2πft) + ε2, where ε1 and ε2 are noise with variances of 0.5 and 0.3, respectively. Kalman filtering estimates the true state through state transition matrix A = [10; 01] and observation matrix H = [10; 01], and after iterative update, the fused signal x_f(t) is obtained, with mean square error reduced to 0.1. The fused data sequence is analyzed for vibration frequency by fast Fourier transform (FFT), and the main frequency component is extracted, for example, the main frequency is found to be 52 Hz, indicating a resonance risk of the structure. The stress change rate sequence is calculated by time series difference, and analysis shows that the change rate reaches 11.2 MPa / s at a certain time, exceeding the threshold, indicating a potential fatigue crack. The fusion data is further input into a long short-term memory network (LSTM) with a time step of 20 ms and a hidden layer of 64 units to predict the mechanical response in the next 10 seconds, and the output sequence shows that the stress peak value may reach 15 MPa, combined with the main frequency analysis, revealing that the instability mechanism is resonance and fatigue coupling.

[0083] S106, frequency domain analysis is performed according to the real-time mechanical response data sequence to obtain a frequency spectrum feature of the response data and determine a structural vibration mode under the combined action of wave-current coupling and seismic load.

[0084] The real-time mechanical response data sequence is obtained by a sensor, and a fast Fourier transform algorithm is used to convert the data into a frequency domain to obtain a frequency spectrum feature. If there is a significant low-frequency component in the frequency spectrum feature, a wavelet transform method is used to perform time-frequency analysis on the data sequence to determine a frequency distribution of the wave-current coupling effect. According to the frequency distribution, a periodic component of the wave-current coupling is extracted from the frequency spectrum feature to obtain a structural response characteristic under the action of the wave and the current. By comparing the structural response characteristic under the action of the wave and the current with a preset seismic load spectrum model, the additional influence of the seismic load on the structure is determined. If the additional influence of the seismic load exceeds a preset threshold, a finite element analysis method is used to determine a vibration mode of the structure in combination with the frequency spectrum feature. According to the vibration mode of the structure, modal parameters are obtained, the vibration components under the combined action of the wave-current coupling and the seismic load are decomposed, and a comprehensive vibration characteristic is obtained. By using the modal superposition method, the dynamic response of the structure is calculated based on the comprehensive vibration characteristic to determine the final vibration mode.

[0085] For example, for frequency domain analysis of the structural vibration mode under the combined action of wave-current coupling and seismic load, real-time mechanical response data sequence is first collected, such as vibration data recorded by an acceleration sensor of a certain offshore platform. Assuming that the sampling frequency is 1000 Hz and the time span is 60 seconds, a data sequence of 60000 sampling points is obtained, including acceleration values in x, y, and z directions, with a unit of m / s 2 , and a typical value of 0.1 to 1.5 m / s 2 . The data preprocessing uses a fast Fourier transform (FFT) algorithm to convert the time-domain acceleration signal into a frequency-domain spectrum. Specifically, the MATLAB fft function is used, the input data is 60000 points, the window function is set to Hanning window to reduce spectral leakage, and the window length is consistent with the data length. The calculation result is the spectral amplitude and phase, the frequency resolution is 1 / 60=0.0167 Hz, and the spectral range is 0 to 500 Hz (Nyquist frequency). When analyzing the frequency spectrum feature, significant peak frequencies are identified, such as the amplitudes of 0.8 m / s 2 , 0.6 m / s 2 , and 0.4 m / s 2The peak values of the frequency spectrum indicate the main vibration modes of the structure. Modal analysis is performed on these frequencies, combined with a finite element model (assuming a platform model established on the ANSYS platform, containing 10,000 degrees of freedom), to calculate the natural frequencies of the structure, verify that 2Hz and 5Hz correspond to the lateral and longitudinal modes of the platform, and 10Hz may be related to wave-current coupling excitation. The effect of seismic load is analyzed by inputting a typical seismic wave (such as the El Centro seismic wave, peak acceleration 0.3g, duration 30 seconds) for time history analysis, superimposed with wave-current excitation (assuming a flow speed of 1m / s, wave height of 2m, and period of 8 seconds), and the Newmark-β method is used to solve the structural response, resulting in an increase in energy in the low frequency band (0-5Hz) in the frequency spectrum, indicating that seismic and wave-current coupling amplify low-frequency vibrations. Finally, by comparing the peak values of the frequency spectrum and the natural frequencies, the dominant vibration modes of the structure at 2Hz and 5Hz are determined, and the enhancement effect of seismic and wave-current coupling on the modal is evaluated, and the modal parameter table (frequency, modal shape, damping ratio) is output.

[0086] S107, extract key instability indicators from the structure vibration mode, use a support vector machine classifier to classify the indicators, and determine the instability type and occurrence probability corresponding to the classification result.

[0087] Vibration signal data is obtained from the structure vibration mode, and the signal is decomposed by Fourier transform to obtain frequency components and amplitude characteristics. According to the frequency components and amplitude characteristics, principal component analysis is used to extract key instability indicators and determine the indicator set. Data preprocessing is performed on the indicator set, and the indicator values are normalized by standardization method to obtain the standardized indicator set. If the dimension of the standardized indicator set is higher than the preset threshold, recursive feature elimination algorithm is used for feature selection to obtain the simplified feature set. The simplified feature set is classified and trained by support vector machine to obtain the instability type classification model. The occurrence probability of each instability type is calculated by kernel density estimation method according to the output of the classification model to determine the probability distribution. According to the probability distribution, combined with the preset instability type threshold, the final instability type and occurrence probability are determined.

[0088] Exemplarily, when extracting the key instability indicators from the structural vibration signals, the features can be obtained through time-frequency analysis methods. First, the collected vibration signals (sampling frequency 1000 Hz, duration 10 seconds, amplitude range -5 to 5 mm) are subjected to fast Fourier transform (FFT) to obtain the frequency spectrum, and the main frequency (e.g. 50 Hz), amplitude (e.g. 2.3 mm) and harmonic component (e.g. amplitude 0.8 mm at 100 Hz) are extracted as preliminary features. Then, the wavelet transform (Daubechies4 wavelet is selected, and 4 layers are decomposed) is used to extract the time-frequency features, and the wavelet coefficient energy ratio (such as the 3rd layer coefficient energy accounts for 35.7% of the total energy) is calculated. In order to capture the nonlinear characteristics, the approximate entropy of the signal is calculated (embedding dimension 2, threshold 0.2, and the approximate entropy value is 0.65). These indicators (main frequency, amplitude, harmonic energy, approximate entropy) constitute the feature vector, which is used for subsequent classification. The support vector machine (SVM) classifier is used to process the feature vector, the radial basis function (RBF) kernel is used, the kernel parameter γ = 0.1, the penalty parameter C = 10, and the model is optimized through 5-fold cross-validation. The training data set contains 200 samples (100 stable, 100 unstable, including two types of instability resonance and flutter), and the test set contains 50 samples. The classification result shows that a test sample feature vector [50, 2.3, 0.8, 0.65] is judged as unstable, belonging to the resonance type, and the classification probability is 0.92. The probability calculation is based on the Platt scaling method of SVM, which maps the decision function value to [0, 1]. In order to verify the result, the confusion matrix of the classifier is analyzed, the accuracy is 94%, the recognition rate of the resonance type is 96%, and the recognition rate of the flutter type is 92%. Through feature importance analysis, the main frequency and approximate entropy contribute most to the classification, and the weights are 0.45 and 0.38 respectively. In order to associate the business, the classification result can be connected with the equipment maintenance system to automatically generate maintenance suggestions, such as “resonance risk detected, recommend checking the excitation source at 50 Hz frequency”. The whole process is realized through Python, calling SciPy library to process FFT and wavelet transform, and Sklearn library to realize SVM classification.

[0089] S108, update the load parameters in the initial marine environment database through the classification result to obtain an optimized coupled load data set for subsequent simulation iteration to improve the accuracy of response capture.

[0090] The classification algorithm is used to obtain a classification result from the marine environment data, and an initial load parameter is determined. If the classification result matches a preset threshold, the load parameter in the database is updated to obtain a first load data set. A clustering algorithm is used to analyze the first load data set to obtain a coupling load feature and generate an optimized data set. The optimized data set is processed through iterative simulation to obtain a response capture result. If the response capture result deviates from a target accuracy by more than a preset threshold, the load parameter is adjusted to generate a second load data set. The iterative simulation is repeated according to the second load data set to obtain a final response capture result and determine accuracy improvement. The final response capture result is verified through data processing to obtain an optimized coupling load data set.

[0091] For example, based on the load parameters in the initial marine environment database, the classification result is updated to generate an optimized coupling load data set to improve the accuracy of response capture in subsequent simulation iterations. First, assume that the initial database contains wind load, wave load and current load parameters, such as wind speed 10 m / s, wave height 2.5 m and current speed 0.8 m / s. The K-means clustering algorithm is used to classify historical marine environment data, the number of clusters K is set to 3, the distance of data points to the centroid is calculated based on the Euclidean distance, and the centroid is iteratively updated until convergence is achieved to obtain three types of environmental states: calm (wind speed < 5 m / s, wave height < 1 m), moderate (wind speed 5-15 m / s, wave height 1-3 m), and severe (wind speed > 15 m / s, wave height > 3 m). According to the classification result, the average load parameter of each type of environment is extracted, for example, the wave load in the calm state is 100 kN, and the corresponding parameter in the database is updated. Then, by using the finite element analysis software ANSYS, a coupling load model is constructed based on the updated load parameters, the wind load is input as 500 kN, the wave load is input as 100 kN, and the current load is input as 50 kN, the displacement response of the marine structure is calculated, and the maximum displacement is 0.02 m. The error between the simulation result and the actual monitoring data is compared, assuming that the actual displacement is 0.025 m, and the error is 20%. If the error exceeds 10%, the load parameter is adjusted, for example, the wave load is increased by 10% to 110 kN, and the simulation iteration is performed again until the error is less than 10%. Finally, an optimized coupling load data set is generated, including wind load 510 kN, wave load 110 kN, and current load 55 kN, which is used for subsequent simulation. The entire process is automatically implemented through a Python script, the scikit-learn library is called to perform K-means clustering, the ANSYS APDL script is used to run finite element analysis, and the loop is iterated until the error requirement is met,

[0092] In another aspect, the embodiment also provides an electronic device, including a memory, a processor, and a computing program stored in the memory and executable on the processor, and the processor implements the method when executing the computing program.

[0093] In another aspect, the present embodiment also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method.

[0094] The above merely provides the preferred embodiments of the present application, and the protection scope of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A simulation method for studying the instability of a submarine data cabin structure under coupled wave-current and seismic loads, characterized in that, include: A three-dimensional structural model of the seabed data cabin was constructed using the finite element analysis method, and initial wave and current parameters and seismic waveform data were obtained from a pre-set marine environment database to generate a coupled load dataset. Computational fluid dynamics simulation is performed based on the coupled load dataset to obtain a dynamic wave-current coupled field. Seismic waveform data is integrated during the simulation to form a nonlinear mechanical effect distribution. When the velocity distribution in the coupled load dataset exceeds a preset threshold, an iterative optimization algorithm is used to adjust the wave height parameter and the seismic intensity coefficient to obtain stable wave-current coupled field simulation results. The structural force vector is extracted from the simulation results of the stable wave-current coupling field. The displacement and stress response of the seabed data cabin in a dynamic environment are predicted using a neural network model, and it is determined whether the instability risk threshold is met. If the instability risk threshold is met, multi-sensor simulation signals are fused to generate a real-time mechanical response data sequence. Frequency domain analysis is performed on the real-time mechanical response data sequence to extract spectral features and determine the structural vibration mode; Key instability indicators are extracted from the vibration modes of the structure, and a classification algorithm is used to process them to obtain the instability type and probability of occurrence. The load parameters in the marine environment database are updated based on the classification results to obtain the optimized coupled load dataset.

2. The method according to claim 1, characterized in that, The construction of the three-dimensional structural model includes: A geometric model of the seabed data cabin was established using the finite element method, and mesh generation and material properties were set. Wave flow parameters and seismic waveform data are obtained from the database, and a standardized parameter set is obtained through data cleaning. Fluid dynamics simulation and spectral analysis were performed on the standardized parameter set to obtain velocity distribution and vibration frequency data, and the coupled load dataset was generated.

3. The method according to claim 1, characterized in that, The computational fluid dynamics simulation includes: Extract the time-series features of the seismic waveform data to obtain the vibration signal input; The vibration signal is superimposed on the wave-current coupling field to generate superimposed flow field data; The flow direction change and the distribution of nonlinear mechanical effects are determined based on the superimposed flow field data.

4. The method according to claim 1, characterized in that, The iterative optimization algorithm includes: When the flow velocity distribution exceeds the threshold, flow velocity anomaly points are extracted and a flow velocity anomaly dataset is generated. Based on the aforementioned velocity anomaly dataset, an optimization algorithm is used to adjust the wave height parameter and seismic intensity coefficient; The stable wave-current coupling field simulation results are generated based on the adjusted parameters.

5. The method according to claim 1, characterized in that, The predicted displacement and stress response include: The force vector of the structure is decomposed to obtain the force vector components; The neural network model is trained using the force vector components as input data. The predicted values ​​of displacement response and stress response are calculated using the neural network model.

6. The method according to claim 1, characterized in that, The fused multi-sensor analog signals include: The vibration frequency and stress change signals collected by the sensor are preprocessed to generate a denoised signal sequence; The denoised signal sequence is fused using a weighted average method to form a unified multidimensional fused data sequence.

7. The method according to claim 1, characterized in that, The extracted key instability indicators include: Perform a Fourier transform on the vibration mode signal to extract frequency components and amplitude features; Key indicators were extracted using principal component analysis to form an indicator set. The index set is classified using a support vector machine classifier to obtain the instability type and probability.

8. The method according to claim 1, characterized in that, The updated load parameters include: Determine the initial load parameters based on the classification results; Cluster analysis is performed on the load parameters to generate an optimized dataset; The final optimized coupled load dataset is obtained by iteratively simulating the optimized dataset.

9. An electronic device comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, characterized in that, When the processor executes the computing program, it implements the method of any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-8.

Images