A method for evaluating wave-induced pore pressure response of seafloor sediments based on real sea conditions

Abstract

The present invention relates to the field of marine engineering technology, and specifically to a method for evaluating the wave-induced pore pressure response of seabed sediments based on real sea conditions, comprising the following steps: S1: collecting basic physical parameters of seabed sediments; S2: collecting measured data of ocean wave conditions; S3: establishing a sediment pore pressure response model through a recursive neural network based on the data collected by S1 and S2; S4: simulating the wave-induced pore pressure response to obtain the pore pressure response curve of the sediment under different wave conditions; S5: evaluating the pore pressure change of seabed sediments under different wave conditions according to the pore pressure response curve simulated by S4. The present invention, by combining recursive neural networks and principal component analysis, accurately simulates and evaluates the pore pressure change of seabed sediments under different wave conditions and the soil deformation caused by it, and provides scientific and reliable technical support for marine engineering design and sediment stability analysis.

Description

Technical Field

[0001] The invention relates to the field of marine engineering technology, and in particular to a method for evaluating wave-induced pore pressure response of seabed sediments based on real sea conditions. Background Art

[0002] With the continuous development of marine engineering, the stability of seabed sediments has become one of the important factors affecting the safety of marine structures; especially under the action of waves, the changes in pore pressure of seabed sediments and the resulting soil deformation have an important impact on the stability of sediments and the safety of engineering structures; the pore pressure of seabed sediments refers to the pressure of water in the soil, which will change under the action of external forces such as waves, thereby affecting the mechanical properties, deformation and stability of the sediments; traditionally, the technology for studying the pore pressure response of seabed sediments mainly relies on physical experiments and theoretical models, but these methods are often difficult to accurately simulate the behavior of sediments under complex sea conditions, and the calculation process is cumbersome, and it is impossible to provide real-time feedback on the dynamic interaction between waves and sediments.

[0003] Existing methods for analyzing pore pressure of seabed sediments usually rely on single experimental data or limited numerical simulations, and are unable to provide real-time and accurate predictions of pore pressure changes in actual marine engineering projects. At the same time, existing technologies are mostly limited to experimental results under certain specific conditions, and lack a comprehensive assessment of sediment pore pressure changes under complex ocean wave conditions. This makes it difficult for existing technologies to accurately predict the changes in sediment pore pressure under different wave conditions and their impact on soil stability when faced with the complex interactions between ocean waves and sediments. Summary of the invention

[0004] Based on the above objectives, the present invention provides a method for evaluating the wave-induced pore pressure response of seabed sediments based on real sea conditions.

[0005] A method for evaluating wave-induced pore pressure response of seafloor sediments based on real sea conditions comprises the following steps:

[0006] S1: Collect basic physical parameters of seafloor sediments, including soil density, porosity and shear modulus;

[0007] S2: Collect measured data on ocean wave conditions, including wave height, wave period, and wave speed;

[0008] S3: Based on the data collected by S1 and S2, a sediment pore pressure response model is established through a recursive neural network to predict the change of sediment pore pressure under wave action;

[0009] S4: Input the measured data of wave conditions collected by S2 and the basic physical parameters of seabed sediments collected by S1 into the sediment pore pressure response model to simulate the wave-induced pore pressure response, and obtain the pore pressure response curves of sediments under different wave conditions;

[0010] S5: Based on the pore pressure response curve obtained by S4 simulation, the pore pressure changes of seabed sediments under different wave conditions are evaluated, and the soil deformation and stability caused by pore pressure are analyzed.

[0011] Optionally, the S1 specifically includes:

[0012] S11: Obtain soil samples from seafloor sediments using a gravity sampler, measure the mass of the sample by weighing, and calculate the soil density based on the sample volume;

[0013] S12: The collected soil samples were processed using the gas expansion method, and the porosity was calculated by measuring the volume change of the samples under standard conditions;

[0014] S13: Use a static triaxial test device to apply different stress levels to the soil sample, record the stress-strain relationship curve, and determine the shear modulus by analyzing the slope of the curve.

[0015] Optionally, the S2 specifically includes:

[0016] S21: Deploy measurement equipment in the sea area through buoys or ocean wave detectors to monitor the rise and fall of waves in real time and record the maximum height from the peak to the trough of the wave as wave height data;

[0017] S22: Use a sea surface buoy or a wave period detector to monitor the waves and record the time interval between two adjacent wave crests or wave troughs as wave period data;

[0018] S23: Use a multi-point sensor array arranged along the wave propagation direction to calculate the wave speed by measuring the arrival time of the wave crest or trough at different locations

[0019] Optionally, the S3 specifically includes:

[0020] S31: Preprocessing the basic physical parameter data of seafloor sediments collected by S1 and the measured data of ocean wave conditions collected by S2, including removing outliers in the data, filling in missing data items, and converting all data into a unified standard scale;

[0021] S32: Perform principal component analysis on the preprocessed soil density, porosity, shear modulus, wave height, wave period and wave velocity data, and extract the first three principal components with the highest explained variance ratio as the extracted features;

[0022] S33: Design the network structure of the recursive neural network, specifically setting the number of hidden layers to two, the number of neurons in each layer to 128, and the ReLU activation function; and select the long short-term memory network as a variant of the recursive neural network;

[0023] S34: Divide the data set processed by S31 and S32 into a training set and a validation set, specifically using 70% of the data as the training set and 30% of the data as the validation set; use the training set to train the LSTM model, and optimize the model parameters through the back propagation algorithm; during the training process, use the early stopping method to monitor the loss function of the validation set to prevent the model from overfitting;

[0024] S35: Use the validation set to evaluate the trained LSTM network model, calculate the mean square error and mean absolute error between the model prediction results and the actual observation values; and adjust the model's hyperparameters based on the error indicators to obtain the final sediment pore pressure response model.

[0025] Optionally, the S32 specifically includes:

[0026] S321: Calculate the covariance matrix of the preprocessed data set. The covariance matrix is ​​used to describe the linear correlation between the data features. The calculation formula is: ,in, is the covariance matrix, is the i-th data sample, is the mean vector of sample data, n is the number of samples;

[0027] S322: Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors by solving the following characteristic equation: ,in, is the covariance matrix, is the eigenvalue, v is the corresponding eigenvector;

[0028] S323: sort all eigenvalues ​​from large to small, and select the eigenvectors corresponding to the three largest eigenvalues ​​after sorting as the main components to be extracted;

[0029] S324: Calculate the variance contribution rate of each principal component. The formula is: Variance contribution rate ,in, is the eigenvalue of the i-th principal component, It is the sum of all eigenvalues, and the variance contribution rate indicates the proportion of the original data variance explained by the principal component;

[0030] S325: Select the first three principal components with the highest variance contribution as the extracted features.

[0031] Optionally, the S35 includes:

[0032] S351: Use the validation set to evaluate the trained LSTM model, input the feature data in the validation set into the trained LSTM model, and obtain the model's predicted output ;

[0033] S352: Calculate the mean square error between the model prediction results and the actual observation values. The formula is: ,in, is the number of samples in the validation set, is the predicted value of the i-th sample, is the actual observed value of the i-th sample, and MSE is the mean square error;

[0034] S353: Calculate the mean absolute error between the model prediction results and the actual observations. The formula is: , where MAE is the mean absolute error;

[0035] S354: According to the calculated mean square error and mean absolute error, the hyperparameters of the LSTM model are adjusted, and the optimal hyperparameters are determined through multiple experiments using a grid search or random search optimization method;

[0036] S355: Retrain the LSTM model according to the optimized hyperparameters and evaluate the performance of the model again until the mean square error and mean absolute error reach the predetermined accuracy requirements, and finally obtain the optimal sediment pore pressure response model, which is expressed as: ,in, is the sediment pore pressure response predicted by the LSTM model, W and b are the weight and bias in the LSTM model, and X is the input feature.

[0037] Optionally, the S4 specifically includes:

[0038] S41: align the wave condition data and sediment physical parameters in time series to ensure that the data at each moment correspond to each other, and convert the aligned data into a format suitable for input of the recurrent neural network model, including a normalized numerical matrix or tensor structure;

[0039] S42: inputting the unified data set processed in S41 into the sediment pore pressure response model established in step S3;

[0040] S43: forward propagating the input wave conditions and sediment physical parameter data, calculating the output according to the weight and bias parameters learned during training, generating the pore pressure response prediction value corresponding to each batch of input data, and forming the pore pressure response time series data;

[0041] S44: Plot the pore pressure response data of the time series in S43 into a curve graph to show the trend of pore pressure changes with time or wave conditions.

[0042] Optionally, the S43 specifically includes:

[0043] Input layer: receives processed data ;

[0044] Hidden layer: For each time step t, the recurrent neural network calculates the hidden state at the current time t , and its calculation formula is: ,in, is the hidden state at the previous moment, is the weight matrix of the hidden layer, is the weight matrix from the input layer to the hidden layer, is the current input data, is the bias term of the hidden layer, is the activation function;

[0045] Output layer: After the hidden layer calculates the hidden state at the current moment, the output is calculated using weighted summation and activation function , predict the pore pressure response value, the calculation formula is: ,in, is the predicted pore pressure response, is the weight matrix of the output layer, is the bias term of the output layer;

[0046] The output for all time steps t Collect and form complete pore pressure response time series data , where T is the total number of time steps.

[0047] Optionally, the S5 specifically includes:

[0048] S51: extracting the maximum pore pressure value, the minimum pore pressure value and the amplitude of pore pressure change under different wave conditions according to the pore pressure response curve generated in S4;

[0049] S52: Based on the maximum pore pressure, minimum pore pressure and pore pressure variation amplitude extracted in S51, the time domain characteristics of the pore pressure are calculated, including the average value, standard deviation and peak value of the pore pressure, and the pore pressure variation characteristics of the seabed sediments under different wave conditions are obtained;

[0050] S53: According to the pore pressure change characteristics, the soil deformation analysis model based on the effective stress method is used to calculate the soil deformation under different wave conditions by combining the pore pressure change data and the mechanical properties of the soil. The calculation formula of the soil deformation is: ,in, is the deformation of the soil, is the effective stress under pore pressure, is the initial effective stress, E is the elastic modulus of the soil;

[0051] S54: Analyze the stability of the sediment based on the calculated soil deformation and use the soil stability assessment formula to evaluate whether the seabed sediment will become unstable under different wave conditions.

[0052] Optionally, the stability evaluation formula is: , where S is the safety factor of the soil, is the shear strength of the soil, is the normal stress; if , it indicates the risk of instability of seabed sediments.

[0053] Beneficial effects of the present invention:

[0054] The present invention effectively processes sediment physical parameters and wave condition data through recursive neural network and principal component analysis technology, overcoming the limitations of traditional methods under complex sea conditions. By establishing a sediment pore pressure response model, real-time monitoring and evaluation of sediment pore pressure changes can be achieved, thereby providing a more scientific basis for marine engineering design.

[0055] The present invention can comprehensively evaluate the mechanical response and deformation behavior of sediments under different wave conditions by deeply analyzing the soil deformation and stability caused by pore pressure, and provide more accurate technical support for sediment stability prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0057] Figure 1 Schematic diagram of a method for evaluating wave-induced pore pressure response of seafloor sediments according to an embodiment of the present invention;

[0058] Figure 2 It is a schematic diagram of the process of establishing a sediment pore pressure response model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0060] like Figure 1-Figure 2As shown, a method for evaluating the wave-induced pore pressure response of seafloor sediments based on real sea conditions includes the following steps:

[0061] S1: Collect basic physical parameters of seafloor sediments, including soil density, porosity and shear modulus;

[0062] S2: Collect measured data on ocean wave conditions, including wave height, wave period, and wave speed;

[0063] S3: Based on the data collected by S1 and S2, a sediment pore pressure response model is established through a recurrent neural network (RNN). The sediment pore pressure response model is used to predict the change of sediment pore pressure under wave action by considering the relationship between wave load and sediment characteristics;

[0064] S4: Input the measured data of wave conditions collected by S2 and the basic physical parameters of seabed sediments collected by S1 into the sediment pore pressure response model to simulate the wave-induced pore pressure response, and obtain the pore pressure response curves of sediments under different wave conditions;

[0065] S5: Based on the pore pressure response curve obtained by S4 simulation, the pore pressure changes of seabed sediments under different wave conditions are evaluated, and the soil deformation and stability caused by pore pressure are analyzed.

[0066] S1 specifically includes:

[0067] S11: Use a gravity sampler to obtain soil samples from seafloor sediments, measure the mass of the samples by weighing, and calculate the soil density based on the sample volume to ensure that the collected soil density data is highly accurate and representative; the calculation formula for soil density is: ,in, is the soil density (kg / m³), m is the mass of the soil sample (kg), and V is the volume of the sample (m³);

[0068] S12: The collected soil samples are processed by gas expansion method. The porosity is calculated by measuring the volume change of the samples under standard conditions to ensure the accuracy and reliability of the porosity data. The porosity calculation formula is: ,in, is the porosity (%), is the volume of pores in the sample (m³), is the total volume of the sample (m³);

[0069] S13: Use static triaxial test equipment to apply different stress levels to soil samples, record the stress-strain curve, and determine the shear modulus by analyzing the slope of the curve to ensure the accuracy of the shear modulus measurement results; the calculation formula for the shear modulus is: , where G is the shear modulus (Pa), is the shear stress (Pa), is the shear strain (dimensionless); the above steps S11 to S13 respectively use a gravity sampler, a gas expansion method and a static triaxial test equipment to ensure that the measurement of soil density, porosity and shear modulus is of high precision and consistency, providing more reliable technical support for marine engineering.

[0070] S2 specifically includes:

[0071] S21: Deploy measurement equipment in the sea area through buoys or ocean wave detectors to monitor the rise and fall of waves in real time and record the maximum height from the peak to the valley of the wave as wave height data; specifically, the measurement of wave height is achieved through wave sensors installed on buoys on the sea surface, and the wave height calculation formula is: , where H is the wave height (m), is the maximum height of the wave (m), is the minimum height of the wave (m);

[0072] S22: Use sea surface buoys or wave period detectors to monitor waves and record the time interval between two adjacent wave crests or wave troughs as wave period data; the wave period is calculated as follows: , where T is the wave period (seconds), is the time interval between peaks (seconds), N is the number of peaks;

[0073] S23: Use a multi-point sensor array to be arranged along the wave propagation direction, and calculate the wave speed by measuring the arrival time of the wave crest or trough at different positions; the wave speed calculation formula is: , where C is the wave speed (m / s), D is the distance of wave propagation (m), and t is the time required for wave propagation (s); the above steps S21 to S23 respectively measure the wave height, wave period and wave speed through technical equipment such as buoys, wave period detectors and multi-point sensor arrays to ensure that the obtained wave data are highly accurate and representative. Through these detailed measurement methods, the present invention can provide more real and accurate ocean wave condition data, thereby providing more reliable input data for the simulation and evaluation of wave-induced pore pressure response of seabed sediments.

[0074] S3 specifically includes:

[0075] S31: Preprocess the basic physical parameter data of seabed sediments collected by S1 and the measured data of ocean wave conditions collected by S2, including removing outliers in the data, filling in missing data items, and converting all data into a unified standard scale to ensure data consistency and comparability;

[0076] S32: Perform principal component analysis on the pre-processed soil density, porosity, shear modulus, wave height, wave period and wave velocity data, and extract the first three principal components with the highest explained variance ratio as the extracted features; ensure that the extracted key features can fully reflect the influence of wave conditions and sediment physical parameters on pore pressure response;

[0077] S33: Design the network structure of the recurrent neural network (RNN), specifically setting the number of hidden layers to two, the number of neurons in each layer to 128, and the ReLU activation function; and select the long short-term memory network (LSTM) as a variant of the recurrent neural network to enhance the model's ability to capture time series data;

[0078] S34: Divide the data set processed by S31 and S32 into a training set and a validation set, specifically using 70% of the data as the training set and 30% of the data as the validation set; use the training set to train the LSTM model and optimize the model parameters through the back propagation algorithm; during the training process, use the Early Stopping method to monitor the loss function of the validation set to prevent the model from overfitting and ensure the generalization ability of the model;

[0079] S35: Use the validation set to evaluate the trained LSTM network model, and calculate the mean square error (MSE) and mean absolute error (MAE) between the model prediction results and the actual observation values; and adjust the model's hyperparameters, such as learning rate, batch size, etc., according to the error indicators to obtain the final sediment pore pressure response model; the above steps establish a high-precision sediment pore pressure response model through specific steps such as data preprocessing, feature selection, recursive neural network structure construction, model training, verification and optimization. This method uses the powerful learning ability of recursive neural networks to effectively capture the complex relationship between wave conditions and sediment characteristics, thereby improving the accuracy and reliability of pore pressure response prediction. As a result, it is possible to more accurately evaluate the changes in pore pressure of seabed sediments under different sea conditions and their impact on soil stability, providing solid technical support for marine engineering design and safety assessment.

[0080] S32 specifically includes:

[0081] S321: Calculate the covariance matrix of the preprocessed data set. The covariance matrix is ​​used to describe the linear correlation between the data features. The calculation formula is: ,in, is the covariance matrix, is the i-th data sample, is the mean vector of sample data, n is the number of samples;

[0082] S322: Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors by solving the following characteristic equation: ,in, is the covariance matrix, is the eigenvalue, v is the corresponding eigenvector;

[0083] S323: sort all eigenvalues ​​from large to small, and select the eigenvectors corresponding to the three largest eigenvalues ​​after sorting as the extracted principal components; these eigenvectors are the features after principal component analysis, which can explain the variance of the data to the greatest extent;

[0084] S324: Calculate the variance contribution rate of each principal component. The formula is: Variance contribution rate ,in, is the eigenvalue of the i-th principal component, It is the sum of all eigenvalues, and the variance contribution rate indicates the proportion of the original data variance explained by the principal component;

[0085] S325: Select the top three principal components with the highest variance contribution rate as the extracted features for subsequent model training; the above steps can accurately extract the top three principal components that can best explain the variance in the data set through the calculation of the covariance matrix, the solution of eigenvalues ​​and eigenvectors, and the selection of variance contribution rates. This feature extraction method effectively reduces the data dimension and can capture the most representative features, thereby improving the effect of subsequent model training.

[0086] S35 includes:

[0087] S351: Use the validation set to evaluate the trained long short-term memory network (LSTM) model, input the feature data in the validation set into the trained LSTM model, and obtain the model's predicted output ;

[0088] S352: Calculate the mean square error between the model prediction results and the actual observation values. The formula is: ,in, is the number of samples in the validation set, is the predicted value of the i-th sample, is the actual observed value of the i-th sample, and MSE is the mean square error, which is used to measure the error size of the model prediction;

[0089] S353: Calculate the mean absolute error between the model prediction results and the actual observations. The formula is: , where MAE is the mean absolute error, which measures the average level of absolute error between the model prediction value and the actual value;

[0090] S354: According to the calculated mean square error and mean absolute error, adjust the hyperparameters of the LSTM model (such as learning rate, number of neuron layers, number of neurons in each layer, training cycle, etc.), use grid search or random search optimization method, and determine the best hyperparameters through multiple experiments;

[0091] S355: Retrain the LSTM model according to the optimized hyperparameters and evaluate the performance of the model again until the mean square error and mean absolute error reach the predetermined accuracy requirements, and finally obtain the optimal sediment pore pressure response model, which is expressed as: ,in, is the sediment pore pressure response predicted by the LSTM model, W and b are the weight and bias in the LSTM model respectively, and X is the input feature (including soil density, porosity, shear modulus, wave height, wave period and wave speed data). The model obtains the optimal weight and bias through training, and can accurately predict the changes in sediment pore pressure under different sea conditions; the above steps can comprehensively measure the accuracy of model prediction by using mean square error (MSE) and mean absolute error (MAE) as evaluation indicators, and improve the prediction accuracy of the model by optimizing hyperparameters; the final LSTM model not only has a high prediction ability, but also can adapt to different sea conditions, provide reliable sediment pore pressure response prediction results, and provide strong support for marine engineering design and soil stability analysis.

[0092] S4 specifically includes:

[0093] S41: align the wave condition data and sediment physical parameters in time series to ensure that the data at each moment correspond to each other, and convert the aligned data into a format suitable for input of the recurrent neural network model, including a normalized numerical matrix or tensor structure;

[0094] S42: inputting the unified data set processed in S41 into the sediment pore pressure response model established in step S3;

[0095] S43: forward propagating the input wave conditions and sediment physical parameter data, calculating the output according to the weight and bias parameters learned during training, generating the pore pressure response prediction value corresponding to each batch of input data, and forming the pore pressure response time series data;

[0096] S44: Plot the pore pressure response data of the time series in S43 into a graph to show the trend of pore pressure changes with time or wave conditions; the above steps and methods can efficiently process and analyze large-scale ocean data, generate clear pore pressure response curves, and provide intuitive and reliable data support for marine engineering design and sediment stability assessment.

[0097] The forward propagation in S43 specifically includes:

[0098] Input layer: receives processed data ;

[0099] Hidden layer: For each time step t, the recurrent neural network calculates the hidden state at the current time t , and its calculation formula is: ,in, is the hidden state at the previous moment, is the weight matrix of the hidden layer, is the weight matrix from the input layer to the hidden layer, is the current input data, is the bias term of the hidden layer, is an activation function, such as ReLU or sigmoid function;

[0100] Output layer: After the hidden layer calculates the hidden state at the current moment, the output is calculated using weighted summation and activation function , predict the pore pressure response value, the calculation formula is: ,in, is the predicted pore pressure response, is the weight matrix of the output layer, is the bias term of the output layer;

[0101] The output for all time steps t Collect and form complete pore pressure response time series data , where T is the total number of time steps; the above steps finally generate the pore pressure response prediction value of the corresponding time step through the hidden layer state calculation of the recursive neural network and the prediction of the output layer, forming the pore pressure response time series data. This process not only improves the prediction accuracy of the model, but also effectively visualizes the pore pressure response time series, which is convenient for decision-making and evaluation in actual engineering.

[0102] S5 specifically includes:

[0103] S51: According to the pore pressure response curve generated in S4, the maximum pore pressure value, the minimum pore pressure value and the amplitude of pore pressure change under different wave conditions are extracted to evaluate the pore pressure fluctuation of seabed sediments under the action of waves;

[0104] S52: Based on the maximum pore pressure, minimum pore pressure and pore pressure variation amplitude extracted in S51, the time domain characteristics of the pore pressure are calculated, including the average value, standard deviation and peak value of the pore pressure, and the pore pressure variation characteristics of the seabed sediments under different wave conditions are obtained;

[0105] S53: According to the pore pressure change characteristics, the soil deformation analysis model based on the effective stress method is used to calculate the soil deformation under different wave conditions by combining the pore pressure change data and the mechanical properties of the soil. The calculation formula of the soil deformation is: ,in, is the deformation of the soil, is the effective stress under pore pressure, is the initial effective stress, and E is the elastic modulus of the soil. This formula calculates the deformation of the soil under the action of waves based on the pore pressure change and the mechanical properties of the soil.

[0106] S54: Analyze the stability of the sediment based on the calculated soil deformation and use the soil stability assessment formula to evaluate whether the seabed sediment will become unstable under different wave conditions.

[0107] The stability evaluation formula is: , where S is the safety factor of the soil, is the shear strength of the soil, is the normal stress; if , it indicates that the seabed sediment is at risk of instability; through the above steps S51 to S54, by extracting the pore pressure fluctuation characteristics, calculating the time domain statistics, and using the effective stress method to analyze the soil deformation and evaluate the stability, a comprehensive evaluation of the pore pressure change on the soil stability is ensured; this method can accurately reflect the dynamic changes of sediment pore pressure under different wave conditions and its impact on soil stability, providing a scientific basis for marine engineering design and sediment stability evaluation.

[0108] In order to verify the effectiveness of the present invention "A method for evaluating wave-induced pore pressure response of seafloor sediments based on real sea conditions", the following experimental test steps were designed and implemented, and relevant experimental data were collected to support the verification results;

[0109] Experimental purpose: To verify the accuracy and reliability of the above method in evaluating the pore pressure changes of seabed sediments and soil stability under actual sea conditions, to ensure that the model can accurately predict the pore pressure response under different wave conditions and effectively evaluate the deformation and stability of sediments;

[0110] Experimental steps and related data

[0111] 1. Data Collection

[0112] Seabed sediment physical parameter collection

[0113] A sea area was selected as the experimental area, and a gravity sampler was used to collect seabed sediment samples, and the soil density, porosity and shear modulus were measured. The specific operations and experimental data are as follows:

[0114] Soil density collection

[0115] Sampling point A: 1800 kg / m³

[0116] Sampling point B: 1850 kg / m³

[0117] Sampling point C: 1780 kg / m³

[0118] Porosity determination (S12)

[0119] Sampling point A: 35%

[0120] Sampling point B: 33%

[0121] Sampling point C: 36%

[0122] Shear modulus measurement

[0123] Sampling point A: 1500 Pa

[0124] Sampling point B: 1550 Pa

[0125] Sampling point C: 1480 Pa

[0126] Ocean wave condition field data collection

[0127] Wave sensors were deployed in the experimental area to monitor wave height, wave period and wave speed in real time and collect continuous wave data. The specific operations and experimental data are as follows:

[0128] wave height

[0129] Time point 1: 2.5 m

[0130] Time point 2: 3.0 m

[0131] Time point 3: 2.8 m

[0132] Wave period

[0133] Time point 1: 8 seconds

[0134] Time point 2: 10 seconds

[0135] Time point 3: 9 seconds

[0136] Wave speed

[0137] Time point 1: 5 m / s

[0138] Time point 2: 6 m / s

[0139] Time point 3: 5.5 m / s

[0140] 2. Model building and training

[0141] Data preprocessing

[0142] The collected data on seabed sediment physical parameters and wave conditions were cleaned and standardized; the processed data are shown in Table 1:

[0143] Table 1. Data on physical parameters of seafloor sediments and ocean wave conditions collected

[0144] Time Soil density (kg / m³) Porosity(%) Shear modulus(Pa) Wave height(m) Wave period (seconds) Wave speed (m / s) 1 1800 35 1500 2.5 8 5 2 1850 33 1550 3 10 6 3 1780 36 1480 2.8 9 5.5

[0145] Feature selection and extraction

[0146] Use principal component analysis to reduce the dimension of the preprocessed data and extract the first three principal components with the highest explained variance ratio as key features. The PCA results are as follows:

[0147] Principal component 1 (PC1): explained variance 40%;

[0148] Principal component 2 (PC2): explained variance 25%;

[0149] Principal component 3 (PC3): explained variance 20%;

[0150] Recurrent Neural Network Model Construction and Training

[0151] A long short-term memory network (LSTM) model with two hidden layers, 128 neurons in each layer and ReLU activation function was constructed; the processed data set was divided into a 70% training set (time points 1-2) and a 30% validation set (time point 3). The model parameters were optimized by the back propagation algorithm, and the early stopping method was used to prevent overfitting; the following error indicators were recorded during the training process:

[0152] Training set error (MSE): 0.004

[0153] Validation set error (MSE): 0.005

[0154] Training set error (MAE): 0.015

[0155] Validation set error (MAE): 0.02

[0156] 3. Simulation of wave-induced pore pressure response

[0157] Data integration and input: The wave condition data and sediment physical parameters are aligned in time series and converted into a format suitable for LSTM model input, as shown in Table 2:

[0158] Table 2 Key feature data extracted by principal component analysis

[0159] Time PC1 PC2 PC3 1 0.8 0.6 0.5 2 0.85 0.55 0.6 3 0.78 0.65 0.55

[0160] Forward propagation and response generation: Through the forward propagation calculation of the model, the pore pressure response prediction value corresponding to each wave condition is generated to form the pore pressure response time series data, which is plotted into a curve graph;

[0161] Predicted pore pressure response (time point 3): 1.2 MPa;

[0162] Pore ​​pressure response time series graph: shows the trend of pore pressure change with time and wave conditions;

[0163] 4. Pore pressure change and stability assessment

[0164] Analysis of pore pressure changes: extract the maximum pore pressure, minimum pore pressure and pore pressure change amplitude in the pore pressure response curve, and calculate the average value, standard deviation and peak value of the pore pressure;

[0165] Maximum pore pressure: 1.5 MPa

[0166] Minimum pore pressure: 1.0 MPa

[0167] Pore ​​pressure variation: 0.5 MPa

[0168] Average pore pressure: 1.2 MPa

[0169] Standard deviation: 0.2 MPa

[0170] Peak value: 1.5 MPa

[0171] Soil deformation and stability analysis: Use the effective stress method to calculate soil deformation and evaluate the stability of sediments based on the safety factor formula;

[0172] The deformation calculation formula is: ;

[0173] The stability evaluation formula is: ;because , indicating that the seafloor sediments are at risk of instability;

[0174] Experimental results: By processing and analyzing the data collected in the actual sea area in the above steps, the following results are obtained:

[0175] Model prediction accuracy: The mean square error (MSE) of the model on the validation set is 0.005, and the mean absolute error (MAE) is 0.02, indicating that the model has high prediction accuracy;

[0176] Pore ​​pressure response curve: The generated pore pressure response curve is highly consistent with the actual observation data and can accurately reflect the dynamic change trend of pore pressure under different wave conditions;

[0177] Stability assessment: By calculating the safety factor, it was found that the safety factor of sediments under extreme wave conditions was 0.83, indicating the risk of instability.

[0178] The conclusion is as follows:

[0179] The above experimental results verify the effectiveness and reliability of the method of the present invention in practical applications. By combining recursive neural networks and principal component analysis technology, it can accurately simulate and predict the changes in pore pressure of seabed sediments under different wave conditions, and effectively evaluate its impact on soil stability. This method overcomes the limitations of traditional methods under complex sea conditions and provides scientific and reliable technical support for marine engineering design and sediment stability assessment.

[0180] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for evaluating wave-induced pore pressure response of seafloor sediments based on real sea conditions, characterized in that: The following steps are involved: S1: Collect basic physical parameters of seafloor sediments, including soil density, porosity and shear modulus; S2: Collect measured data on ocean wave conditions, including wave height, wave period, and wave speed; S3: Based on the data collected by S1 and S2, a sediment pore pressure response model is established through a recursive neural network to predict the change of sediment pore pressure under wave action, including: S31: Preprocessing the basic physical parameter data of seafloor sediments collected by S1 and the measured data of ocean wave conditions collected by S2, including removing outliers in the data, filling in missing data items, and converting all data into a unified standard scale; S32: Perform principal component analysis on the preprocessed soil density, porosity, shear modulus, wave height, wave period and wave velocity data, and extract the first three principal components with the highest explained variance ratio as the extracted features; S33: Design the network structure of the recursive neural network, specifically setting the number of hidden layers to two, the number of neurons in each layer to 128, and the ReLU activation function; and select the long short-term memory network as a variant of the recursive neural network; S34: Divide the data set processed by S31 and S32 into a training set and a validation set, specifically using 70% of the data as the training set and 30% of the data as the validation set; use the training set to train the LSTM model, and optimize the model parameters through the back propagation algorithm; during the training process, use the early stopping method to monitor the loss function of the validation set to prevent the model from overfitting; S35: Use the validation set to evaluate the trained LSTM network model, calculate the mean square error and mean absolute error between the model prediction results and the actual observation values; and adjust the model's hyperparameters based on the error indicators to obtain the final sediment pore pressure response model; S4: Input the measured data of wave conditions collected by S2 and the basic physical parameters of seabed sediments collected by S1 into the sediment pore pressure response model to simulate the wave-induced pore pressure response, and obtain the pore pressure response curves of sediments under different wave conditions; S5: Based on the pore pressure response curve obtained by S4 simulation, the pore pressure changes of seabed sediments under different wave conditions are evaluated, and the soil deformation and stability caused by pore pressure are analyzed.

2. The method for evaluating the wave-induced pore pressure response of seabed sediments based on real sea conditions according to claim 1, characterized in that: The S1 specifically includes: S11: Obtain soil samples from seafloor sediments using a gravity sampler, measure the mass of the sample by weighing, and calculate the soil density based on the sample volume; S12: The collected soil samples were processed using the gas expansion method, and the porosity was calculated by measuring the volume change of the samples under standard conditions; S13: Use a static triaxial test device to apply different stress levels to the soil sample, record the stress-strain relationship curve, and determine the shear modulus by analyzing the slope of the curve.

3. The method for evaluating the wave-induced pore pressure response of seabed sediments based on real sea conditions according to claim 1, characterized in that: The S2 specifically includes: S21: Deploy measurement equipment in the sea area through buoys or ocean wave detectors to monitor the rise and fall of waves in real time and record the maximum height from the peak to the trough of the wave as wave height data; S22: Use a sea surface buoy or a wave period detector to monitor the waves and record the time interval between two adjacent wave crests or wave troughs as wave period data; S23: Use a multi-point sensor array arranged along the wave propagation direction to calculate the wave speed by measuring the arrival time of the wave crest or trough at different locations.

4. The method for evaluating the wave-induced pore pressure response of seafloor sediments based on real sea conditions according to claim 1, characterized in that: The S32 specifically includes: S321: Calculate the covariance matrix of the preprocessed data set. The covariance matrix is ​​used to describe the linear correlation between the data features. The calculation formula is: ,in, is the covariance matrix, is the i-th data sample, is the mean vector of sample data, n is the number of samples; S322: Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, and obtain the eigenvalues ​​and corresponding eigenvectors by solving the following characteristic equation: ,in, is the covariance matrix, is the eigenvalue, v is the corresponding eigenvector; S323: sort all eigenvalues ​​from large to small, and select the eigenvectors corresponding to the three largest eigenvalues ​​after sorting as the main components to be extracted; S324: Calculate the variance contribution rate of each principal component. The formula is: Variance contribution rate ,in, is the eigenvalue of the i-th principal component, It is the sum of all eigenvalues, and the variance contribution rate indicates the proportion of the original data variance explained by the principal component; S325: Select the first three principal components with the highest variance contribution as the extracted features.

5. The method for evaluating wave-induced pore pressure response of seabed sediments based on real sea conditions according to claim 4, characterized in that: The S35 tool comprises: S351: Use the validation set to evaluate the trained LSTM model, input the feature data in the validation set into the trained LSTM model, and obtain the model's predicted output ; S352: Calculate the mean square error between the model prediction results and the actual observation values. The formula is: ,in, is the number of samples in the validation set, is the predicted value of the i-th sample, is the actual observed value of the i-th sample, and MSE is the mean square error; S353: Calculate the mean absolute error between the model prediction results and the actual observations. The formula is: , where MAE is the mean absolute error; S354: According to the calculated mean square error and mean absolute error, the hyperparameters of the LSTM model are adjusted, and the optimal hyperparameters are determined through multiple experiments using a grid search or random search optimization method; S355: Retrain the LSTM model according to the optimized hyperparameters and evaluate the performance of the model again until the mean square error and mean absolute error reach the predetermined accuracy requirements, and finally obtain the optimal sediment pore pressure response model, which is expressed as: ,in, is the sediment pore pressure response predicted by the LSTM model, W and b are the weight and bias in the LSTM model, and X is the input feature.

6. The method for evaluating wave-induced pore pressure response of seafloor sediments based on real sea conditions according to claim 1, characterized in that: The S4 specifically includes: S41: align the wave condition data and sediment physical parameters in time series to ensure that the data at each moment correspond to each other, and convert the aligned data into a format suitable for input of the recurrent neural network model, including a normalized numerical matrix or tensor structure; S42: inputting the unified data set processed in S41 into the sediment pore pressure response model established in step S3; S43: forward propagating the input wave conditions and sediment physical parameter data, calculating the output according to the weight and bias parameters learned during training, generating the pore pressure response prediction value corresponding to each batch of input data, and forming the pore pressure response time series data; S44: Plot the pore pressure response data of the time series in S43 into a curve graph to show the trend of pore pressure changes with time or wave conditions.

7. The method for evaluating wave-induced pore pressure response of seabed sediments based on real sea conditions according to claim 6, characterized in that: The S43 specifically includes: Input layer: receives processed data ; Hidden layer: For each time step t, the recurrent neural network calculates the hidden state at the current time t , and its calculation formula is: ,in, is the hidden state at the previous moment, is the weight matrix of the hidden layer, is the weight matrix from the input layer to the hidden layer, is the current input data, is the bias term of the hidden layer, is the activation function; Output layer: After the hidden layer calculates the hidden state at the current moment, the output is calculated using weighted summation and activation function , predict the pore pressure response value, the calculation formula is: ,in, is the predicted pore pressure response, is the weight matrix of the output layer, is the bias term of the output layer; The output for all time steps t Collect and form complete pore pressure response time series data , where T is the total number of time steps.

8. The method for evaluating wave-induced pore pressure response of seabed sediments based on real sea conditions according to claim 1, characterized in that: The S5 specifically includes: S51: extracting the maximum pore pressure value, the minimum pore pressure value and the amplitude of pore pressure change under different wave conditions according to the pore pressure response curve generated in S4; S52: Based on the maximum pore pressure, minimum pore pressure and pore pressure variation amplitude extracted in S51, the time domain characteristics of the pore pressure are calculated, including the average value, standard deviation and peak value of the pore pressure, and the pore pressure variation characteristics of the seabed sediments under different wave conditions are obtained; S53: According to the pore pressure change characteristics, the soil deformation analysis model based on the effective stress method is used to calculate the soil deformation under different wave conditions by combining the pore pressure change data and the mechanical properties of the soil. The calculation formula of the soil deformation is: ,in, is the deformation of the soil, is the effective stress under pore pressure, is the initial effective stress, E is the elastic modulus of the soil; S54: Analyze the stability of the sediment based on the calculated soil deformation and use the soil stability assessment formula to evaluate whether the seabed sediment will become unstable under different wave conditions.

9. The method for evaluating wave-induced pore pressure response of seafloor sediments based on real sea conditions according to claim 1, characterized in that: The stability evaluation formula is: , where S is the safety factor of the soil, is the shear strength of the soil, is the normal stress; if , it indicates the risk of instability of seabed sediments.

Images