A Multi-Scenario Adaptive Wave Simulation Method and System Based on Fluid Dynamics

Abstract

This invention discloses a multi-scenario adaptive wave simulation method based on fluid dynamics, comprising the following steps: Step 1: wave feature sensing; Step 2: multi-scale coupled quantization; Step 3: collaborative adaptive decision-making; and Step 4: adaptive simulation execution. This invention also discloses a multi-scenario adaptive wave simulation system based on fluid dynamics. This invention introduces a pre-trained decision model, upgrading the mapping from wave feature parameters to SPH simulation parameters from rule-based weighted calculation to learning-based intelligent mapping, enabling continuous evolution of parameter decision-making capabilities. It constructs a complete technical closed loop of "wave feature perception—multi-scale coupled quantization—collaborative adaptive decision-making—adaptive simulation execution," achieving breakthroughs in spatial resolution adaptation, accurate quantization of multi-scale nonlinear disturbances, and intelligent parameter decision-making.

Description

Technical Field

[0001] This invention relates to the field of numerical simulation technology in fluid dynamics, and more specifically to a multi-scenario adaptive wave simulation method and system based on fluid dynamics. Background Technology

[0002] Wave impact is one of the main environmental loads faced by marine engineering structures (such as offshore platforms, breakwaters, ships, and offshore wind turbine foundations) during their service life. Accurately simulating the impact boundary forces of waves on structures is of vital engineering significance for structural safety assessment, fatigue life prediction, and optimized design.

[0003] In numerical simulation methods, smoothed particle fluid dynamics, as a pure Lagrangian, meshless particle method, has been widely used in wave impact problems in recent years. Compared with the traditional finite element method or finite volume method, smoothed particle fluid dynamics has significant advantages: it not only eliminates the need to generate complex styles, naturally avoiding mesh distortion caused by large deformations, but also efficiently captures strongly nonlinear free surface flow phenomena such as wave breaking, water splashing, and gas-liquid mixing. Furthermore, the handling of fluid-structure interaction interfaces is relatively direct, requiring no special interface tracking algorithms. Therefore, smoothed particle fluid dynamics is considered a powerful numerical tool for studying the interaction mechanism between waves and structures. However, in the actual marine environment, waves exhibit an extremely wide scale distribution, ranging from millimeter-scale capillary waves and meter-scale nearshore breaking waves to tens or even hundreds of meters of typhoon waves and tsunami waves. The physical characteristics, energy transfer mechanisms, and impact modes on structures of waves at different scales vary greatly.

[0004] Existing technologies have disclosed attempts at adaptive smooth particle hydrodynamics methods, such as Chinese patent: A wave impact boundary force simulation and analysis method based on SPH, authorized announcement number: CN120706330B. This invention improves the adaptability of multi-scale wave impact simulation by calculating the wave divergence index and modal coupling coefficient, and dynamically adjusting the critical time step and artificial viscous system in SPH simulation.

[0005] However, this method still has the following limitations and shortcomings in application:

[0006] One drawback is that this method only adjusts the time step and the artificial viscous system, but the spatial distribution of particles remains uniform or fixed. For areas with drastic changes such as local high pressure zones and fracture zones caused by small-scale wave impacts, if the particles are not pre-encrypted, it is impossible to capture the key spatial pressure gradient. However, full-area encryption would result in a large waste of computational resources.

[0007] Second shortcoming: Although the method proposes a modal coupling coefficient, it is mainly based on time-frequency domain correlation analysis and fails to accurately locate the region with the strongest multi-scale wave nonlinear interference (i.e., the "interference-sensitive region") from the spatial topology of the pressure field, resulting in insufficient spatial targeting of parameter adjustment.

[0008] The third shortcoming is that the method uses a rule-based weighted calculation formula to adjust the time step and viscosity coefficient. The coupling relationship between the parameters (such as the mutual constraints between spatial resolution, time step and viscosity coefficient) cannot be explicitly modeled and optimized, making it difficult to achieve the globally optimal parameter configuration in complex scenarios.

[0009] To address the shortcomings of the prior art, this invention provides a multi-scenario adaptive wave simulation method and system based on fluid dynamics, thereby solving the aforementioned technical problems. Summary of the Invention

[0010] This invention discloses a multi-scenario adaptive wave simulation method and system based on fluid dynamics. Its main purpose is to address the shortcomings and deficiencies of existing technologies. By constructing a closed-loop control framework that includes wave feature perception, multi-scale coupling quantization, and adaptive decision-making of simulation parameters, it dynamically optimizes the spatial resolution, time step, and artificial viscosity coefficient in existing SPH simulations, thereby achieving accurate and efficient simulation of waves at different scales and their coupling effects.

[0011] The technical solution adopted in this invention is as follows:

[0012] A multi-scenario adaptive wave simulation method based on fluid dynamics includes the following specific steps:

[0013] Step 1: Sensing wave characteristics: Acquire spatiotemporal evolution data of the target wave, calculate the energy divergence index to characterize the spatial dispersion characteristics of wave energy, and the modal energy entropy to characterize the frequency domain complexity of wave-structure coupling.

[0014] Step 2: Multi-scale coupling quantization step: Based on the topological structure of the pressure field on the structure surface, identify the spatial region with the strongest multi-scale wave nonlinear interference, which is denoted as the interference sensitive region, and calculate the modal coupling coefficient that characterizes the intensity of nonlinear interference between waves of different scales within this region.

[0015] Step 3: Collaborative Adaptive Decision Making: Input the energy divergence index, modal energy entropy, and simulation coupling system into the pre-trained decision model. The model will collaboratively make decisions and output three core parameters for smooth particle hydrodynamics simulation: namely, spatial resolution distribution function, optimized critical duration, and optimized artificial viscosity coefficient.

[0016] Step 4: Adaptive simulation execution steps: Configure the particle system using the spatial resolution distribution function obtained from the decision in Step 3, then control the time progression using the optimized critical duration, and finally update the inter-particle interaction forces using the optimized artificial viscosity coefficient to perform the simulation analysis of wave impact boundary forces.

[0017] Furthermore, the calculation method for the energy dissipation index includes the following specific steps:

[0018] 1) Obtain a time-series image sequence of waves from the start of rolling to the breaking of the wave;

[0019] 2) Determine the wave initiation and breaking moments based on the fractal dimension change rate of the image boundary;

[0020] 3) Calculate the wave height ratio between the start-up time and the breakage time as the cumulative energy value;

[0021] 4) Calculate the ratio of the fractal dimension at the moment of breakage to the time it takes for the wave to reach the structure after breaking as an indicator of complex potential energy.

[0022] 5) Calculate the ratio of the wave height at the moment of breakage to the average wave height of the same period as the probability of abnormal waves;

[0023] 6) The product of the energy accumulation value, the complex potential energy index and the probability of the abnormal wave is used as the energy divergence index of the wave.

[0024] Furthermore, the identification method for the interference-sensitive area in step two includes the following specific steps:

[0025] 1) Spatial interpolation is performed on the pressure signals at various test points on the surface of the structure to generate a continuous spatial pressure field;

[0026] 2) Take the numerical gradient of the spatial pressure field to obtain the pressure gradient value;

[0027] 3) Calculate the Poincaré index at each location in the pressure gradient field, and denote the points where the index is non-zero as force field singularities;

[0028] 4) Perform spatial clustering on the force field singularities, select the cluster family containing the most singularities, and denote the region enclosed by its convex hull boundary as the interference sensitive area.

[0029] Furthermore, the calculation method for the modal coupling coefficient in step two includes the following specific steps:

[0030] 1) Adaptive decomposition of the coupled pressure signal in the interference-sensitive area yields multiple component signals of different scales;

[0031] 2) Select the two component signals with the largest frequency difference as the target signals, and calculate the wave coherence coefficient between them;

[0032] 3) Calculate the energy contribution values ​​of the two target signals, and calculate the energy coupling ratio based on the linear superposition and synthesis signal of the two;

[0033] 4) The product of the wave coherent system and the energy coupling ratio is used as the simulation coupling coefficient.

[0034] Furthermore, the decision model in step three is a pre-trained neural network model, and its training process includes the following steps:

[0035] 1) Construct a training dataset. Each training data point contains the energy divergence index, modal energy entropy, and modal coupling coefficient of the sample wave as input features, and the optimal combination of simulation parameters obtained through network search or Bayesian optimization as labels.

[0036] 2) The neural network is trained under supervised learning with the goal of minimizing the error between the simulation results and the actual physical observations.

[0037] 3) The trained neural network can directly map wave feature parameters to the co-optimized spatial resolution, time step, and artificial viscosity coefficient.

[0038] Furthermore, the spatial resolution distribution function in step three is a non-uniform distribution function, and its constraints include: Condition 1: In spatial regions where the energy divergence index is higher than the first threshold and / or the modal coupling coefficient is higher than the second threshold, the particle density is not lower than the preset fine resolution threshold; Condition 2: In other regions, the particle density is not higher than the preset sparse resolution threshold.

[0039] Furthermore, the decision-making method for optimizing the critical time in step three includes the following specific steps:

[0040] 1) Obtain the baseline critical time step;

[0041] 2) Calculate the coefficient of variation of the force-structure coupling factor for all waves within the scale cluster to which the wave belongs. The force-structure coupling factor is the ratio of modal energy entropy to structural response hysteresis coefficient.

[0042] 3) The product of the coefficient of variation and the energy divergence index is negatively correlated to obtain the time adjustment coefficient;

[0043] 4) The product of the baseline critical time step and the time adjustment coefficient is used as the optimization critical time.

[0044] Furthermore, the decision-making method for optimizing the artificial viscosity coefficient in step three includes the following steps:

[0045] 1) Obtain the baseline artificial viscosity coefficient for the wave scale;

[0046] 2) Calculate the ratio of the optimization critical time step to the baseline critical time step, and remember the time scaling factor;

[0047] 3) The product of the duration scaling factor and the modal coupling coefficient is used as the viscosity correction coefficient;

[0048] 4) The sum of the baseline artificial viscosity coefficient and the viscosity correction coefficient is used as the optimized artificial viscosity coefficient.

[0049] A multi-scenario adaptive wave simulation system based on fluid dynamics is disclosed. The simulation system includes a data acquisition module, a feature calculation module, a decision module, and a simulation execution module. The data acquisition module is used to acquire wave spatiotemporal evolution data and pressure and application signals on the surface of structures. The feature calculation module is used to execute wave feature perception steps and multi-scale coupled quantization steps. The decision module is built into a pre-trained decision model and is used to execute collaborative adaptive decision steps. The simulation execution module is used to execute adaptive simulation execution steps.

[0050] As can be seen from the above description and explanation of the present invention, compared with the prior art, the advantages of the present invention are as follows:

[0051] Advantage 1: This invention focuses computational resources on regions of intense physical processes by adapting spatial resolution. Compared with fixed high-resolution methods, the computational efficiency can be improved by 1-2 orders of magnitude. By adapting the time step, it avoids numerical divergence in small-scale scenes and invalid computation in large-scale scenes.

[0052] Advantage 2: This invention introduces modal coupling coefficients within the framework of smooth particle hydrodynamics to quantify the nonlinear disturbances of multi-scale waves, and dynamically adjusts the artificial viscosity accordingly, enabling the simulation to accurately reproduce complex physical phenomena such as large waves modulating small waves and energy cascading. The accuracy of impact load prediction is significantly higher than that of the traditional linear superposition method.

[0053] Advantage 3: This invention introduces a pre-trained decision model, which improves the mapping from wave feature parameters to SPH simulation parameters from rule-based weighted calculation to learning-based intelligent mapping, enabling the parameter decision-making capability to continuously evolve. It constructs a complete technical closed loop of "wave feature perception - multi-scale coupled quantization - collaborative adaptive decision-making - adaptive simulation execution", achieving breakthrough progress in spatial resolution adaptation, accurate quantization of multi-scale nonlinear disturbances, and intelligent parameter decision-making. Detailed Implementation

[0054] The specific embodiments of the present invention will be further described and explained below.

[0055] A multi-scenario adaptive wave simulation method based on fluid dynamics includes the following specific steps:

[0056] Step 1: Sensing wave characteristics: Acquire spatiotemporal evolution data of the target wave, calculate the energy divergence index to characterize the spatial dispersion characteristics of wave energy, and the modal energy entropy to characterize the frequency domain complexity of wave-structure coupling.

[0057] Step 2: Multi-scale coupling quantization step: Based on the topological structure of the pressure field on the structure surface, identify the spatial region with the strongest multi-scale wave nonlinear interference, which is denoted as the interference sensitive region, and calculate the modal coupling coefficient that characterizes the intensity of nonlinear interference between waves of different scales within this region.

[0058] Step 3: Collaborative Adaptive Decision Making: Input the energy divergence index, modal energy entropy, and simulation coupling system into the pre-trained decision model. The model will collaboratively make decisions and output three core parameters for smooth particle hydrodynamics simulation: namely, spatial resolution distribution function, optimized critical duration, and optimized artificial viscosity coefficient.

[0059] Step 4: Adaptive simulation execution steps: Configure the particle system using the spatial resolution distribution function obtained from the decision in Step 3, then control the time progression using the optimized critical duration, and finally update the inter-particle interaction forces using the optimized artificial viscosity coefficient to perform the simulation analysis of wave impact boundary forces.

[0060] Furthermore, the calculation method for the energy dissipation index includes the following specific steps:

[0061] 1) Obtain a time-series image sequence of waves from the start of rolling to the breaking of the wave;

[0062] 2) Determine the wave initiation and breaking moments based on the fractal dimension change rate of the image boundary;

[0063] 3) Calculate the wave height ratio between the start-up time and the breakage time as the cumulative energy value;

[0064] 4) Calculate the ratio of the fractal dimension at the moment of breakage to the time it takes for the wave to reach the structure after breaking as an indicator of complex potential energy.

[0065] 5) Calculate the ratio of the wave height at the moment of breakage to the average wave height of the same period as the probability of abnormal waves;

[0066] 6) The product of the energy accumulation value, the complex potential energy index and the probability of the abnormal wave is used as the energy divergence index of the wave.

[0067] Furthermore, the identification method for the interference-sensitive area in step two includes the following specific steps:

[0068] 1) Spatial interpolation is performed on the pressure signals at various test points on the surface of the structure to generate a continuous spatial pressure field;

[0069] 2) Take the numerical gradient of the spatial pressure field to obtain the pressure gradient value;

[0070] 3) Calculate the Poincaré index at each location in the pressure gradient field, and denote the points where the index is non-zero as force field singularities;

[0071] 4) Perform spatial clustering on the force field singularities, select the cluster family containing the most singularities, and denote the region enclosed by its convex hull boundary as the interference sensitive area.

[0072] Furthermore, the calculation method for the modal coupling coefficient in step two includes the following specific steps:

[0073] 1) Adaptive decomposition of the coupled pressure signal in the interference-sensitive area yields multiple component signals of different scales;

[0074] 2) Select the two component signals with the largest frequency difference as the target signals, and calculate the wave coherence coefficient between them;

[0075] 3) Calculate the energy contribution values ​​of the two target signals, and calculate the energy coupling ratio based on the linear superposition and synthesis signal of the two;

[0076] 4) The product of the wave coherent system and the energy coupling ratio is used as the simulation coupling coefficient.

[0077] Furthermore, the decision model in step three is a pre-trained neural network model, and its training process includes the following steps:

[0078] 1) Construct a training dataset. Each training data point contains the energy divergence index, modal energy entropy, and modal coupling coefficient of the sample wave as input features, and the optimal combination of simulation parameters obtained through network search or Bayesian optimization as labels.

[0079] 2) The neural network is trained under supervised learning with the goal of minimizing the error between the simulation results and the actual physical observations.

[0080] 3) The trained neural network can directly map wave feature parameters to the co-optimized spatial resolution, time step, and artificial viscosity coefficient.

[0081] Furthermore, the spatial resolution distribution function in step three is a non-uniform distribution function, and its constraints include: Condition 1: In spatial regions where the energy divergence index is higher than the first threshold and / or the modal coupling coefficient is higher than the second threshold, the particle density is not lower than the preset fine resolution threshold; Condition 2: In other regions, the particle density is not higher than the preset sparse resolution threshold.

[0082] Furthermore, the decision-making method for optimizing the critical time in step three includes the following specific steps:

[0083] 1) Obtain the baseline critical time step;

[0084] 2) Calculate the coefficient of variation of the force-structure coupling factor for all waves within the scale cluster to which the wave belongs. The force-structure coupling factor is the ratio of modal energy entropy to structural response hysteresis coefficient.

[0085] 3) The product of the coefficient of variation and the energy divergence index is negatively correlated to obtain the time adjustment coefficient;

[0086] 4) The product of the baseline critical time step and the time adjustment coefficient is used as the optimization critical time.

[0087] Furthermore, the decision-making method for optimizing the artificial viscosity coefficient in step three includes the following steps:

[0088] 1) Obtain the baseline artificial viscosity coefficient for the wave scale;

[0089] 2) Calculate the ratio of the optimization critical time step to the baseline critical time step, and remember the time scaling factor;

[0090] 3) The product of the duration scaling factor and the modal coupling coefficient is used as the viscosity correction coefficient;

[0091] 4) The sum of the baseline artificial viscosity coefficient and the viscosity correction coefficient is used as the optimized artificial viscosity coefficient.

[0092] A multi-scenario adaptive wave simulation system based on fluid dynamics is disclosed. The simulation system includes a data acquisition module, a feature calculation module, a decision module, and a simulation execution module. The data acquisition module is used to acquire wave spatiotemporal evolution data and pressure and application signals on the surface of structures. The feature calculation module is used to execute wave feature perception steps and multi-scale coupled quantization steps. The decision module is built into a pre-trained decision model and is used to execute collaborative adaptive decision steps. The simulation execution module is used to execute adaptive simulation execution steps.

[0093] Example 1

[0094] A multi-scenario adaptive wave simulation method based on fluid dynamics is proposed, which can be applied to the impact load analysis of nearshore structures (such as breakwaters and offshore wind turbine foundations) under typhoon wave action. The method includes the following steps:

[0095] Step 1: Wave Feature Sensing: Deploy a wave monitoring system in the target sea area. This system includes an array of wave height meters positioned in front of the structure to collect free surface elevation data; an array of miniature pressure sensors installed on the surface of the structure to collect wave impact pressure time history signals; and a synchronously triggered high-speed camera to capture images of wave free surface evolution.

[0096] 1. Calculation of energy divergence index:

[0097] 1) Extract the complete wave process from the start to the break of waves from a high-speed camera image sequence. Use the Canny edge detection algorithm to extract the wave boundary lines of each frame. Calculate the separation dimension D of the boundary lines using box counting. f (t).

[0098] 2) Calculate the rate of change of fractal dimension between adjacent frames.

[0099] r(t) = [D f (t+Δt)-D f (t)] / Δt, smooth the rate of change curve, and find the time corresponding to the first maximum point, which is recorded as the starting time t. curl The time corresponding to the global peak point is recorded as the breaking time t. break .

[0100] 3) Read the wave height H at the start of winding from the wave height meter data. curl and the wave height H at the moment of breakage break Calculate the cumulative energy value: E acc =H break / H curl .

[0101] 4) Calculate the fractal dimension D at the moment of breakup. f (t) break ), and the time T it takes for the wave to travel from the breakage point to the surface of the structure. travel (Distance divided by wave speed) Calculate the complex potential energy index: C pot =D f (t break ) / T travel .

[0102] 5) Calculate the probability of a deformed wave: Calculate the wave height H at the moment the wave breaks. break Average wave height at all wave breaking moments during the same period H break The ratio of M rogue =H break / H break .

[0103] 6) Finally, the energy divergence index D is calculated as: D = E acc ×C pot ×M rogue .

[0104] 2. Modal energy entropy calculation

[0105] 1) Average the pressure signals from all pressure sensors on the structure surface under a single wave impact to obtain the overall pressure signal P. global (t).

[0106] 2) For P global (t) Perform three-level wavelet packet decomposition to obtain wavelet packet decomposition coefficients for 8 frequency bands. For the i-th frequency band, calculate its energy characterization value: Where C i,j Let be the coefficients of the i-th wavelet packet.

[0107] 3) Calculate the energy percentage for each frequency band: .

[0108] 4) Modal energy entropy H mode The calculation is as follows: .

[0109] The larger the value, the wider the wave energy distribution frequency band, and the smaller or more disordered the wave scale; the smaller the value, the more concentrated the wave energy is near the dominant frequency, and the larger or more regular the wave scale.

[0110] 3. Calculation of structural response hysteresis coefficient

[0111] 1) Extract the overall strain signal S global (t).

[0112] 2) Find all the maximum points of the overall pressure signal, and record the corresponding time as t. p,1 , t p,2 …t p,M Find all the maximum points of the overall strain signal, and record the corresponding time as t. s,1 , t s,2 …t s,N .

[0113] 3) For each time t when the pressure reaches its maximum value P，m Find the first time t after which the strain reaches its maximum value. S，n (satisfying t) S，n >t P，m ), calculate the time interval Tm = t S，n -t P，m The average deformation hysteresis time is obtained by averaging over all m values. T .

[0114] 4) Calculate the wave period T wave : The average time interval between all adjacent maximum points on the overall pressure signal.

[0115] 5) Lag coefficient L delay = T / T wave .

[0116] Step 2: Multi-scale Coupling Quantization

[0117] 1. Identification of interference-sensitive areas

[0118] 1) Perform radial basis function interpolation on the pressure values ​​of all pressure sensors at the same moment under a single wave impact to generate a spatial pressure field P(x,y,t).

[0119] 2) Calculate the numerical gradient of the pressure field to obtain the pressure gradient field.

[0120] .

[0121] 3) Calculate the Poincaré index at each grid point in the gradient field. The specific method is as follows: Draw a small circle centered at the point, calculate the net rotation angle of the gradient vector around the circle, and divide by 2π. Points with non-zero Poincaré indices (usually ±1 / 2, ±1, etc.) are denoted as force field singularities.

[0122] 4) Perform DBSCAN clustering based on Euclidean distance on all force field singularities (radius parameter ε = 0.2m, minimum number of points MinPts = 5). Select the cluster containing the most singularities, calculate the convex hull boundary of the cluster, and the area enclosed by the convex hull is the disturbance sensitive area of ​​the wave.

[0123] 2. Calculation of modal coupling coefficients

[0124] 1) Extract the average value of all pressure sensor signals within the interference-sensitive area, and denot it as the coupled pressure signal P. coup (t).

[0125] 2) Use variational mode decomposition (VMD) to transform P coup (t) is decomposed into K intrinsic mode components (IMFs). k (t) (In this embodiment, K = 5).

[0126] 3) Prepare three single-scale wave reference signals in advance: small-scale reference signal Ref s (t)(wave height 0.3m, period 1.5s), mesoscale reference signal Ref m (t)(wave height 1.2m, period 4.0s), large-scale reference signal Ref l (t)(wave height 3.5m, period 8.0s). Calculate the mutual information entropy between each IMF component and the three reference signals. If any of the normalized components exceeds the threshold of 0.3, then the IMF is marked as the analysis signal.

[0127] 4) Calculate the energy contribution value of each analyzed signal:

[0128]

[0129] 5) Select the two analysis signals with the largest frequency difference, denoted as IMFa(t) and IMFb(t), and calculate the wave coherence coefficient between them.

[0130] First, perform continuous wavelet transform on both signals to obtain the wavelet coefficients.

[0131] ,

[0132] Then calculate the cross wavelet spectrum. ,

[0133] Then calculate their respective wavelet power spectra. ,

[0134] The wave coherence coefficient is defined as:

[0135] Where <> represents the time-scale average.

[0136] 6) Calculate the energy coupling ratio:

[0137] First, synthesize the signal Psyn(t) = IMFa(t) + IMFb(t), and then calculate the energy of the synthesized signal. The coupling ratio is obtained as follows: This value reflects the energy deviation of nonlinear superposition relative to linear superposition.

[0138] 7) Modal coupling coefficients: .

[0139] Step 3: Collaborative Adaptive Decision Making

[0140] A three-layer feedforward neural network was used as the decision model, with 3 nodes in the input layer, 12 nodes in the hidden layer, and 3 nodes in the output layer. During training, 500 training samples were collected. Each sample contained wave feature triples and the optimal simulated parameters for that wave obtained through network search or Bayesian optimization. The Adam optimizer was then used with a learning rate of 0.001, and the training run lasted for 300 epochs.

[0141] Step 4: Adaptive Simulation Execution

[0142] Based on the spatial resolution distribution function of the collaborative adaptive decision-making, non-uniformly distributed particles are generated in the computational domain. The particle spacing is approximately 0.025m in the disturbance-sensitive area and recovers to 0.05m at a distance. The total number of particles is about 30% of that of the uniform fine resolution scheme, significantly reducing the computational load. Then, a predictive correction integral scheme is adopted, using the optimization critical duration of the adaptive decision-making at each time step. The particle distribution and wave characteristics are checked every 50 steps, and the decision model is called back to update the parameters if necessary. Then, the SPH standard kernel function is used to calculate particle interactions, followed by boundary condition processing. Finally, the pressure values ​​of each particle on the structure surface are recorded at each time step, generating pressure distribution cloud maps and time history curves. The simulation continues until the wave energy is basically dissipated or the structural response tends to stabilize, completing the simulation output.

[0143] Example Experiment Results: The total number of particles was approximately 2.6 million, the simulation time was approximately 10.5 hours, the error between the peak pressure and the experimental value was less than 8%, and the secondary pressure pulse generated by the superposition of small-scale waves on the peak of large-scale waves was successfully captured. The simulation efficiency was improved by approximately 85% compared with global high resolution, and the accuracy was close to the level of global high resolution.

[0144] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial improvements made to the present invention using this concept should be considered as infringing on the protection scope of the present invention.

Claims

1. A multi-scenario adaptive wave simulation method based on fluid dynamics, characterized in that: The specific steps include the following: Step 1: Sensing wave characteristics: Acquire spatiotemporal evolution data of the target wave, calculate the energy divergence index to characterize the spatial dispersion characteristics of wave energy, and the modal energy entropy to characterize the frequency domain complexity of wave-structure coupling. Step 2: Multi-scale coupling quantization step: Based on the topological structure of the pressure field on the structure surface, identify the spatial region with the strongest multi-scale wave nonlinear interference, which is denoted as the interference sensitive region, and calculate the modal coupling coefficient that characterizes the intensity of nonlinear interference between waves of different scales within this region. Step 3: Collaborative Adaptive Decision Making: Input the energy divergence index, modal energy entropy, and simulation coupling system into the pre-trained decision model. The model will collaboratively make decisions and output three core parameters for smooth particle hydrodynamics simulation: namely, spatial resolution distribution function, optimized critical duration, and optimized artificial viscosity coefficient. Step 4: Adaptive simulation execution steps: Configure the particle system using the spatial resolution distribution function obtained from the decision in Step 3, then control the time progression using the optimized critical duration, and finally update the inter-particle interaction forces using the optimized artificial viscosity coefficient to perform the simulation analysis of wave impact boundary forces.

2. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The calculation method for the energy dissipation index includes the following specific steps: 1) Obtain a time-series image sequence of waves from the start of rolling to the breaking of the wave; 2) Determine the wave initiation and breaking moments based on the fractal dimension change rate of the image boundary; 3) Calculate the wave height ratio between the start-up time and the breakage time as the cumulative energy value; 4) Calculate the ratio of the fractal dimension at the moment of breakage to the time it takes for the wave to reach the structure after breaking as an indicator of complex potential energy. 5) Calculate the ratio of the wave height at the moment of breakage to the average wave height of the same period as the probability of abnormal waves; 6) The product of the energy accumulation value, the complex potential energy index and the probability of the abnormal wave is used as the energy divergence index of the wave.

3. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The identification method for the interference-sensitive area in step two includes the following specific steps: 1) Spatial interpolation is performed on the pressure signals at various test points on the surface of the structure to generate a continuous spatial pressure field; 2) Take the numerical gradient of the spatial pressure field to obtain the pressure gradient value; 3) Calculate the Poincaré index at each location in the pressure gradient field, and denote the points where the index is non-zero as force field singularities; 4) Perform spatial clustering on the force field singularities, select the cluster family containing the most singularities, and denote the region enclosed by its convex hull boundary as the interference sensitive area.

4. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The calculation method for the modal coupling coefficient in step two includes the following specific steps: 1) Adaptive decomposition of the coupled pressure signal in the interference-sensitive area yields multiple component signals of different scales; 2) Select the two component signals with the largest frequency difference as the target signals, and calculate the wave coherence coefficient between them; 3) Calculate the energy contribution values ​​of the two target signals, and calculate the energy coupling ratio based on the linear superposition and synthesis signal of the two; 4) The product of the wave coherent system and the energy coupling ratio is used as the simulation coupling coefficient.

5. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The decision model in step three is a pre-trained neural network model, and its training process includes the following steps: 1) Construct a training dataset. Each training data point contains the energy divergence index, modal energy entropy, and modal coupling coefficient of the sample wave as input features, and the optimal combination of simulation parameters obtained through network search or Bayesian optimization as labels. 2) The neural network is trained under supervised learning with the goal of minimizing the error between the simulation results and the actual physical observations. 3) The trained neural network can directly map wave feature parameters to the co-optimized spatial resolution, time step, and artificial viscosity coefficient.

6. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The spatial resolution distribution function in step three is a non-uniform distribution function, and its constraints include: Condition 1: In spatial regions where the energy divergence index is higher than the first threshold and / or the modal coupling coefficient is higher than the second threshold, the particle density is not lower than the preset fine resolution threshold; Condition 2: In other regions, the particle density is not higher than the preset sparse resolution threshold.

7. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The decision-making method for optimizing the critical time in step three includes the following specific steps: 1) Obtain the baseline critical time step; 2) Calculate the coefficient of variation of the force-structure coupling factor for all waves within the scale cluster to which the wave belongs. The force-structure coupling factor is the ratio of modal energy entropy to structural response hysteresis coefficient. 3) The product of the coefficient of variation and the energy divergence index is negatively correlated to obtain the time adjustment coefficient; 4) The product of the baseline critical time step and the time adjustment coefficient is used as the optimization critical time.

8. The multi-scenario adaptive wave simulation method based on fluid dynamics according to claim 1, characterized in that: The decision-making method for optimizing the artificial viscosity coefficient in step three includes the following steps: 1) Obtain the baseline artificial viscosity coefficient for the wave scale; 2) Calculate the ratio of the optimization critical time step to the baseline critical time step, and remember the time scaling factor; 3) The product of the duration scaling factor and the modal coupling coefficient is used as the viscosity correction coefficient; 4) The sum of the baseline artificial viscosity coefficient and the viscosity correction coefficient is used as the optimized artificial viscosity coefficient.

9. A multi-scenario adaptive wave simulation system based on fluid dynamics, characterized in that: The simulation system includes a data acquisition module, a feature calculation module, a decision module, and a simulation execution module. The data acquisition module is used to acquire wave spatiotemporal evolution data and pressure and application signals on the surface of structures. The feature calculation module is used to execute wave feature perception steps and multi-scale coupled quantization steps. The decision module is built into a pre-trained decision model and is used to execute collaborative adaptive decision steps. The simulation execution module is used to execute adaptive simulation execution steps.

Images