A Method for Deploying Marine Three-Dimensional Air Gun Arrays Based on Machine Learning Algorithms

Abstract

This application provides a method for configuring a marine 3D airgun array based on a machine learning algorithm, belonging to the field of material or object detection technology. The steps include: collecting historical input parameters and performing data preprocessing; in the data preprocessing, horizontally concatenating two-dimensional sequence features with single-value scalar features to form a high-dimensional feature vector; constructing an original dataset and using it to build an initial model based on the machine learning algorithm; applying the trained configuration model to the marine 3D airgun array to be configured to output the current configuration parameters. This application achieves high-dimensional feature fusion through cross-type feature concatenation in the preprocessing stage. Combined with machine learning technology, it enables rapid adaptation of 3D airgun array configurations under different sea conditions. After training, it can quickly obtain highly accurate marine 3D airgun array configuration parameters, which not only simplifies the calculation steps and reduces the computational resource requirements, but also has high execution efficiency and accuracy that meets practical engineering needs.

Description

Technical Field

[0001] This application belongs to the field of material or object detection technology, specifically relating to a method for configuring a marine three-dimensional air gun array based on machine learning algorithms. Background Technology

[0002] In the field of marine seismic exploration, 3D airgun arrays play a crucial role as seismic source devices. The working principle of a 3D airgun array is to generate shock waves by releasing high-pressure gas, providing sufficient energy for seismic data acquisition. In practice, the deployment and combination mode and trigger time coding scheme of the 3D airgun array have a significant impact on the exploration results. Different subarray deployment and combination modes and trigger time coding schemes can lead to differences in far-field wavelets and spectra, thus affecting the quality and accuracy of the exploration data.

[0003] However, traditional simulation methods have significant limitations when dealing with these complex combinations. Specifically, due to the numerous factors that need to be considered in configuring a three-dimensional air gun array, conventional simulation methods are often time-consuming and inefficient, making it difficult to meet the urgent needs of on-site construction. Therefore, previous work mainly relied on manual experiments to optimize the combination of the best immersion depth and the best delayed excitation time encoding scheme for the exploration target. While manual experience-based screening can achieve the goal to some extent, it still suffers from low efficiency and is easily affected by human factors, leading to fluctuations in the accuracy and reliability of the results.

[0004] In summary, further improvements are needed to overcome the shortcomings of existing technologies and provide an efficient and accurate parameter configuration method for the optimized design of air gun arrays in marine seismic exploration. Summary of the Invention

[0005] This application addresses the problems existing in the prior art by providing a method for configuring marine three-dimensional air gun arrays based on cross-type feature stitching and machine learning algorithms, thereby achieving rapid and accurate output of marine three-dimensional air gun array configurations to solve the aforementioned problems.

[0006] To achieve the above objectives, the technical solution adopted in this application is as follows:

[0007] This application provides a method for configuring a marine three-dimensional air gun array based on machine learning algorithms, including the following steps:

[0008] Historical input parameters characterizing the wavelet characteristics of the marine three-dimensional air gun array source were collected. The historical input parameters included two-dimensional sequence features and single-valued scalar features.

[0009] The historical input parameters are preprocessed. In this preprocessing, at least the two-dimensional sequence features are interpolated or truncated to unify the data length. The two-dimensional sequence features with the unified data length are then horizontally concatenated with the single-value scalar features to form a high-dimensional feature vector. Preferably, the horizontal concatenation involves arranging the two-dimensional sequence features with the unified data length in the order of wavelet sequence and spectrum sequence, and then sequentially concatenating them with the single-value scalar features to form a 1×N-dimensional high-dimensional feature vector, where N is the sum of the total dimension of the two-dimensional sequence features and the dimension of the single-value scalar features. The single-value scalar features are arranged in the order of initial peak value, peak-to-peak value, wave-bulb ratio, minimum frequency band value, and maximum frequency band value.

[0010] Filter the historical configuration parameters corresponding to the historical input parameters;

[0011] The preprocessed historical input parameters and the corresponding historical configuration parameters are combined to form the original dataset, and the original dataset is divided into a training set and a reserved validation set.

[0012] An initial model is constructed based on a machine learning algorithm; the initial model is trained using the training set to obtain a trained model;

[0013] The trained model is validated and fine-tuned using the reserved validation set to obtain the configured model;

[0014] Collect the current input parameters corresponding to the marine three-dimensional air gun array to be configured; perform data preprocessing on the current input parameters, and input the preprocessed current input parameters into the configuration model to output the current configuration parameters of the marine three-dimensional air gun array to be configured.

[0015] Furthermore, the machine learning algorithm used to construct the initial model is one of the following: XGBoost algorithm, LightGBM algorithm, CatBoost algorithm, random forest algorithm, fully connected deep neural network algorithm, or one-dimensional convolutional neural network algorithm.

[0016] Furthermore, during the training process of the initial model, cross-validation is used for training, including the following steps:

[0017] The training set is divided into z training subsets. Each time, z-1 training subsets are used to train the initial model, and one training subset is used to verify the training effect. This process is repeated z times to complete the cross-validation. z is a positive integer and is greater than or equal to 2.

[0018] Furthermore, during the training process of the initial model, hyperparameter optimization is also performed, which employs either a grid search method or a Bayesian optimization method.

[0019] Furthermore, the two-dimensional sequence features are wavelets and spectrum, and the single-valued scalar features include at least the initial peak value, peak-to-peak value, wave-to-bulb ratio, minimum frequency band value, and maximum frequency band value.

[0020] Furthermore, the historical configuration parameters and the current configuration parameters are the sinking depth of each subarray in the marine three-dimensional air gun array and the delayed excitation time between each subarray.

[0021] Furthermore, when interpolating or truncating the two-dimensional sequence features, the length of the unified two-dimensional sequence features is a preset fixed length;

[0022] For two-dimensional sequence features whose length after interpolation is less than the preset fixed length, the preset value is used to fill the preset fixed length.

[0023] Furthermore, when the original dataset is divided into a training set and a reserved validation set, the ratio of the training set to the reserved validation set is 1:1.

[0024] Further, validating the trained model includes the following steps:

[0025] Calculate the root mean square error and mean absolute error of the verification results;

[0026] Based on the root mean square error and the mean absolute error, the verification label of the configuration model is output using a threshold comparison method.

[0027] Furthermore, after outputting the current configuration parameters of the marine three-dimensional air gun array to be configured, the current configuration parameters are fine-tuned, and the fine-tuning range does not exceed 10% of the values ​​of each parameter in the current configuration parameters.

[0028] Compared with the prior art, this application has the following advantages:

[0029] This application achieves high-dimensional feature fusion by splicing cross-type features in the preprocessing stage, thereby maintaining the correlation between parameters in different fields. Combined with machine learning technology, it can quickly adapt the configuration of three-dimensional gun arrays under different sea conditions. After training, it can quickly obtain highly accurate configuration parameters of marine three-dimensional air gun arrays. Unlike conventional simulation or iterative optimization methods, the method of this application simplifies the calculation steps, reduces the computational resource requirements, and has high execution efficiency and accuracy that meets the needs of actual engineering. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a flowchart of a method in a specific embodiment of this application. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0034] It should also be noted that, unless otherwise specified, the methods used in this application are conventional methods; and the raw materials and equipment used are, unless otherwise specified, conventional commercially available products.

[0035] This application proposes a method for configuring a marine three-dimensional air gun array based on machine learning algorithms, such as... Figure 1 As shown, the specific steps include:

[0036] 1. Historical input parameter collection and historical configuration parameter filtering;

[0037] 1.1 Historical Input Parameter Acquisition: Historical input parameters characterizing the wavelet characteristics of the marine three-dimensional airgun array source were acquired. These historical input parameters include two-dimensional sequence features and single-valued scalar features. The two-dimensional sequence features are wavelets and spectra. Specifically, the wavelets are obtained by denoising the original vibration signals acquired by hydrophones, and the spectra are generated by the wavelets through Fast Fourier Transform (FFT). The single-valued scalar features include at least the initial peak value (amplitude of the first wavelet peak, unit: barm), peak-to-peak value (difference between the positive and negative maximum amplitudes of the wavelet, unit: barm), wave-bubble ratio (the ratio of the energy of the main wavelet pulse to the energy of subsequent bubble pulses, with the energy of the first 10ms of the main pulse and the energy of the bubble pulses from 10ms to 50ms), minimum and maximum frequency band values ​​(the boundary of the frequency range where the spectral energy is ≥ 10% of the maximum energy, unit: Hz).

[0038] It should be noted that the sources of historical input parameters must cover existing data types in the field, including measured data from marine seismic exploration, such as vibration data collected by hydrophones in a 3D exploration of an oil field; and also numerical simulation data, such as simulation data of the finite element method based on wave theory. The sample size of a single batch must be no less than 1,000 sets to ensure the generalization ability of subsequent model training.

[0039] 1.2 Historical Configuration Parameter Screening: Historical configuration parameters corresponding to the above-mentioned historical input parameters are screened. These historical configuration parameters are the sinking depth of each subarray in the marine three-dimensional airgun array and the delay excitation time between each subarray. The sinking depth is typically selected as 5-15 meters, with 6-10 meters preferred for shallow sea exploration. The delay excitation time is selected as 0-5 milliseconds to avoid wavelet interference. The screening criteria are parameter combinations that, after practical verification or simulation optimization, can meet the seismic data quality requirements, such as wide effective bandwidth and high signal-to-noise ratio.

[0040] 2. Historical input parameter preprocessing and original dataset construction;

[0041] 2.1 Historical Input Parameter Preprocessing: Historical input parameters undergo data preprocessing, including the following steps:

[0042] First, interpolation or truncation operations are performed on the two-dimensional sequence features (wavelet, spectrum) to unify the length of the two-dimensional sequence features to a preset fixed length. This preset fixed length can be selected according to the application scenario, and is usually set to 256, which corresponds to a wavelet time series covering 0ms-256ms and a spectrum frequency range covering 0Hz-200Hz. The interpolation method can be selected according to the feature type. For example, linear interpolation is commonly used for wavelet sequences, and cubic spline interpolation is commonly used for the spectrum. If the length of the two-dimensional sequence features after interpolation is still insufficient to the preset fixed length, it is filled to the preset fixed length by a preset value. One example is that the preset value is NaN, which can be replaced by the mean of the same batch of data sequences to avoid missing values ​​affecting model training.

[0043] Next, the two-dimensional sequence features of uniform length are horizontally concatenated with the single-valued scalar features to form a high-dimensional feature vector. The horizontal concatenation involves arranging the two-dimensional sequence features of uniform data length in the order of wavelet sequence and then spectrum sequence, and then sequentially concatenating them with the single-valued scalar features to form a 1×N-dimensional high-dimensional feature vector, where N is the sum of the total dimension of the two-dimensional sequence features and the dimension of the single-valued scalar features. The single-valued scalar features are arranged in the order of initial peak value, peak-to-peak value, wave-bulb ratio, minimum frequency band value, and maximum frequency band value.

[0044] Specifically, taking a single sample with a preset fixed length of 256 as an example, in the two-dimensional sequence features, the wavelet is uniformly interpolated to 256 dimensions (time step 1ms, covering 0-256ms), such as [w1,w2,...,w 256 The spectrum is interpolated to a 256-dimensional sequence (frequency step size 0.78125Hz, covering 0-200Hz), such as [s1,s2,...,s...]. 256 The single-valued scalar is 6-dimensional (initial peak value 82.4 barm, peak-to-peak value 127.5 barm, wavelet-to-bulb ratio 22.5, minimum frequency band value 6.42 Hz, maximum frequency band value 88.12 Hz, wavelet energy 1500 J), such as [p1, p2, ..., p6]. After horizontal splicing, it becomes: [w1, w2, ..., w 256 , s1,s2,...,s 256 [p1, p2, ..., p6], forming a 1×518 dimensional feature vector. Dimensions 1-256 correspond to the amplitude at each time point of the wavelet, dimensions 257-512 correspond to the energy at each frequency point of the spectrum, and dimensions 513-518 correspond to six single-valued scalar parameters. This 1×518 dimensional feature vector includes supplementary scalars such as energy features outside the wavelet / spectrum. Each dimension corresponds to a feature value, with no row / column dimension expansion, only feature dimension stacking. At the multi-sample level, multiple high-dimensional feature vectors from single samples are stacked vertically to form an M×N dimensional feature matrix, where M is the number of samples and N is the feature dimension of a single sample. Thus, the feature vectors of the 1000 samples in the above example are arranged vertically to form a 1000×518 dimensional feature matrix, which can be used as the input dataset for model training.

[0045] 2.2 Construction and Partitioning of the Original Dataset: The preprocessed historical input parameters and corresponding historical configuration parameters form the original dataset. The original dataset is stored in the domain-efficient HDF5 format, facilitating the reading and writing of large-scale sequence data. Preferably, the original dataset is randomly divided into a training set and a reserved validation set in a 1:1 ratio. The reserved validation set is only used for the final model generalization ability evaluation and does not participate in the early training and parameter tuning.

[0046] The training set is , The i-th feature vector with m dimensions is represented in this application as an input parameter containing wavelet, spectrum, initial peak value, peak-to-peak value, wave-bulb ratio, minimum frequency band value, and maximum frequency band value. The label of the i-th sample is represented in this application as the corresponding submersion depth and delayed excitation time encoding scheme.

[0047] 3. Initial model construction and training;

[0048] 3.1 Initial Model Construction: An initial model is constructed based on machine learning algorithms. The machine learning algorithm used is one of the following: XGBoost, LightGBM, CatBoost, Random Forest, Fully Connected Deep Neural Network, or One-Dimensional Convolutional Neural Network. Among them, XGBoost is preferred due to its advantages in processing structured data and complex nonlinear relationships. Its specific implementation follows the improved Gradient Boosting Decision Tree (GBDT) framework, specifically: The XGBoost model achieves the mapping between input features and configuration parameters by integrating multiple CART regression trees. Each tree is trained based on the prediction residuals of the previous tree, and the final output is the sum of the prediction values ​​of all trees. Here, residual = historical true value of configuration parameter - prediction value of the first N trees.

[0049] Specifically, the initial model contains K trees. In this application, K represents the number of CART regression trees allocated to the model based on the amount of training set data. A single tree in the model is defined as follows:

[0050] ;

[0051] The above formula can be simplified to:

[0052] ;

[0053] In the formula, Let be the analytical value of the i-th sample; This is the output of the i-th CART regression tree; Representing the Kth decision tree, calculate the sum of the analytical values ​​of each tree and use it as the final analytical value; The objective function is set to the sum of the analytical values ​​of all K decision trees:

[0054] ;

[0055] In the formula, the objective function obj consists of two parts, the first of which is the loss function. The first term is used to evaluate the loss or error between the model's predicted values ​​and the actual values; the second term is the regularization term. , used to control overfitting.

[0056] 3.2 Initial Model Training: The initial model is trained using the training set. The specific process is as follows:

[0057] First, cross-validation is used for training: the training set is divided into z equal training subsets, where z is a positive integer greater than or equal to 2; in this application, z=5, i.e., five-fold cross-validation, where four training subsets are used to train the initial model each time, and one training subset is used to verify the training effect, and this process is repeated five times to complete the cross-validation. This method can make full use of limited data and reduce the risk of overfitting.

[0058] Hyperparameter optimization is performed concurrently during training, employing either a grid search method or a Bayesian optimization method. Specifically, in the hyperparameter grid and settings, the key hyperparameters that the XGBoost model needs to optimize are defined, and a possible range of values ​​is assigned to them, organized in dictionary form, including:

[0059] learning_rate (learning rate / step size);

[0060] max_depth (maximum depth of the tree);

[0061] min_child_weight (minimum leaf node weight);

[0062] subsample / colsample_bytree (sample / feature sampling ratio);

[0063] reg_alpha, reg_lambda (L1 and L2 regularization parameters);

[0064] Grid search and hyperparameter optimization: A grid search is performed. For each set of hyperparameters, the five-fold cross-validation training described above is executed, and the average performance metric (such as mean squared error) of the five-fold validation is used as the evaluation criterion for the quality of that set of parameters. Finally, the hyperparameter combination with the best average performance is selected as the optimal configuration of the model, and the final XGBoost model is retrained on the entire training set (excluding the independent validation set).

[0065] The training process is based on the gradient boosting algorithm, which iteratively builds multiple decision trees. Each tree learns and corrects the prediction residuals of the previous tree, and finally the prediction results of all trees are weighted and summed as the final output.

[0066] In addition, during training, each sample in the training set... The immersion depth and delayed excitation time encoding scheme corresponding to the input parameters of neutron wave, spectrum, initial peak value, peak-to-peak value, wave-bulb ratio, minimum frequency band value, and maximum frequency band value is used to calculate the first-order gradient of the loss function. and second gradient The formulas are as follows:

[0067] ;

[0068] ;

[0069] In the formula, This is the predicted value for the (s-1)th iteration, in order to achieve efficient optimization of the objective function.

[0070] Model training can terminate if any of the following conditions are met:

[0071] (1) The number of iterations reaches the hyperparameter setting value; where the hyperparameter setting value refers to the preset value of the hyperparameter n_estimators (number of iterations);

[0072] (2) The MSE did not decrease after five consecutive rounds of cross-validation;

[0073] (3) The prediction residual for a single tree is less than 0.01.

[0074] After training is completed, the trained model is obtained. To facilitate subsequent deployment, the trained model can be temporarily stored. After verification, it can be serialized and saved as a configuration model in Joblib format.

[0075] 4. Model Validation and Configuration: Model Acquisition;

[0076] The specific steps for validating the trained model are as follows:

[0077] First, the input parameters of the reserved validation set are processed according to the preprocessing method in step 2.1, and then input into the configuration model to obtain the prediction configuration parameters corresponding to the validation set.

[0078] Secondly, the root mean square error (RMSE) and mean absolute error (MAE) of the verification results are calculated. Based on these two error indices, the verification label of the configuration model is output using a threshold comparison method. Furthermore, the threshold settings refer to the accuracy requirements of the marine three-dimensional air gun array configuration, typically: RMSE < 0.5 and MAE < 0.3 for the immersion depth, and RMSE < 0.1 and MAE < 0.05 for the delayed excitation time.

[0079] The validation labels include qualified or unqualified labels. If the validation result does not meet the threshold, it is necessary to return to step 3.2 to readjust the hyperparameters (e.g., expand the grid search range), supplement with high-quality samples (e.g., add measured data across oil fields), or optimize the preprocessing process (e.g., change the interpolation method) until the threshold is met. If the validation meets the threshold, the training model is fine-tuned based on the validation result, such as adjusting the weights of the leaf nodes of a few trees, to finally obtain the configured model.

[0080] 5. Current configuration parameter output and fine-tuning;

[0081] 5.1 Current Input Parameter Acquisition and Preprocessing: Acquire the current input parameters corresponding to the marine three-dimensional air gun array to be configured. The acquisition device and data type must be consistent with the historical input parameters in step 1.1. Perform the preprocessing method of the historical input parameters in step 2.1 on the current input parameters, and keep it completely consistent with the preprocessing of historical input parameters to ensure that the input feature format matches the input requirements of the configuration model.

[0082] 5.2 Current Configuration Parameter Output: Input the preprocessed current input parameters into the configuration model and output the current configuration parameters of the marine three-dimensional air gun array to be configured, namely the sinking depth of each subarray and the delay excitation time between each subarray.

[0083] 5.3 Fine-tuning of current configuration parameters: After outputting the current configuration parameters of the marine three-dimensional air gun array to be configured, it is necessary to fine-tune the current configuration parameters in combination with the on-site construction conditions, such as seabed topography and equipment load-bearing limitations. The fine-tuning range shall not exceed 10% of the value of each parameter in the current configuration parameters. For example, if the predicted sinking depth is 9 meters, but the thickness of the seabed silt layer measured on site is relatively thick, it can be fine-tuned to 8.5 meters (fine-tuning range 5.6% < 10%) to ensure construction safety and effectiveness. Example

[0084] Taking the preferred scheme of using an air gun array as the seismic source for a certain voyage as an example, the actual input parameters and the actual preferred configuration parameters are shown in Table 1 below:

[0085] Table 1 Actual Input Parameters and Optimal Configuration Parameters

[0086]

[0087] Substituting the input parameters into the configuration model of this embodiment, we obtain the configuration parameters output by the model, namely the immersion depth and the delayed excitation time. The overall accuracy obtained by comparing the configuration parameters output by the model with the actual optimized configuration parameters is shown in Table 2 below:

[0088] Table 2 analyzes the immersion depth, delayed excitation, and overall accuracy.

[0089]

[0090] In this embodiment, the average overall accuracy rate in existing examples reaches 93%, which shows that the method in this embodiment can reach the level of practical application.

[0091] Finally, it should be noted that the above content is only used to illustrate the technical solution of this application, and is not intended to limit the scope of protection of this application. Simple modifications or equivalent substitutions made by those skilled in the art to the technical solution of this application shall not depart from the substance and scope of the technical solution of this application.

Claims

1. A method for configuring a marine three-dimensional air gun array based on machine learning algorithms, characterized in that, Includes the following steps: Historical input parameters characterizing the wavelet characteristics of the marine three-dimensional air gun array source are collected. The historical input parameters include two-dimensional sequence features and single-valued scalar features. The two-dimensional sequence features are wavelets and spectrum, and the single-valued scalar features include at least the initial peak value, peak-to-peak value, wave-bulb ratio, minimum value of the frequency band, and maximum value of the frequency band. The historical input parameters are preprocessed; in the data preprocessing, at least the two-dimensional sequence features are interpolated or truncated to unify the data length, and the two-dimensional sequence features with the unified data length are horizontally concatenated with the single-value scalar features to form a high-dimensional feature vector. Filter the historical configuration parameters corresponding to the historical input parameters; The preprocessed historical input parameters and the corresponding historical configuration parameters are combined to form the original dataset, and the original dataset is divided into a training set and a reserved validation set. An initial model is constructed based on a machine learning algorithm; the initial model is trained using the training set to obtain a trained model; The trained model is validated and fine-tuned using the reserved validation set to obtain the configured model; Collect the current input parameters corresponding to the marine three-dimensional air gun array to be configured; perform the data preprocessing on the current input parameters, and input the preprocessed current input parameters into the configuration model to output the current configuration parameters of the marine three-dimensional air gun array to be configured. The historical configuration parameters and the current configuration parameters are the sinking depth of each subarray in the marine three-dimensional air gun array and the delay excitation time between each subarray.

2. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, The machine learning algorithm used to construct the initial model is one of the following: XGBoost algorithm, LightGBM algorithm, CatBoost algorithm, random forest algorithm, fully connected deep neural network algorithm, or one-dimensional convolutional neural network algorithm.

3. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, During the training of the initial model, cross-validation is used, including the following steps: The training set is divided into z training subsets. Each time, z-1 training subsets are used to train the initial model, and one training subset is used to verify the training effect. This process is repeated z times to complete the cross-validation. z is a positive integer and is greater than or equal to 2.

4. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, During the training of the initial model, hyperparameter optimization is also performed, which employs either a grid search method or a Bayesian optimization method.

5. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, When interpolating or truncating the two-dimensional sequence features, the length of the unified two-dimensional sequence features is a preset fixed length; For two-dimensional sequence features whose length after interpolation is less than the preset fixed length, the preset value is used to fill the preset fixed length.

6. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, When the original dataset is divided into a training set and a reserved validation set, the ratio of the training set to the reserved validation set is 1:

1.

7. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, Validating the trained model includes the following steps: Calculate the root mean square error and mean absolute error of the verification results; Based on the root mean square error and the mean absolute error, the verification label of the configuration model is output using a threshold comparison method.

8. The method for configuring a marine three-dimensional air gun array based on a machine learning algorithm according to claim 1, characterized in that, After outputting the current configuration parameters of the marine three-dimensional air gun array to be configured, the current configuration parameters are further fine-tuned, and the fine-tuning range does not exceed 10% of the values ​​of each parameter in the current configuration parameters.

Images