Microseismic source prediction method, device and equipment based on single sensor and storage medium

Abstract

The embodiment of the invention discloses a micro-seismic source prediction method and device based on a single sensor, equipment and a storage medium, and relates to the technical field of micro-seismic positioning. According to the scheme, the method comprises the steps of inputting microseismic signal feature data collected by a single sensor into a microseismic source prediction model to obtain source position information, and obtaining a source position through the microseismic source prediction model trained according to a data set of the microseismic signal feature data collected by a plurality of single sensors. The limitation that a traditional method depends on accurate wave velocity information and high-precision arrival time pickup is overcome, the position of the seismic source can be predicted only through one sensor, the complexity of micro-seismic seismic source prediction is reduced, the application scene is expanded, the method is suitable for a complex rock mass environment, and the installation and maintenance cost is greatly reduced.

Description

Technical Field

[0001] This application relates to the technical field of microseismic location, and particularly to a microseismic source prediction method, device, equipment and storage medium based on a single sensor. Background Art

[0002] There are microscopic defects inside the rock. Under the action of external loads, the expansion of microcracks causes the rock to undergo plastic deformation, and the stored strain energy is released in the form of elastic waves to generate microseismic phenomena. Predicting the location of microseismic sources can help engineers timely detect stress changes inside the rock, so as to early warn of potential rock fracture or collapse risks.

[0003] In the related art, traditional microseismic source prediction methods use multiple sensors to work together, and locate the source by measuring the arrival time difference and wave velocity information of elastic waves. Traditional methods require accurate wave velocity information and high-precision picking of the arrival time of elastic waves, which have great limitations for complex geological conditions and harsh working environments, and the installation and maintenance costs of multiple sensors are relatively high. Summary of the Invention

[0004] Embodiments of this specification provide a microseismic source prediction method based on a single sensor to solve the problems in the prior art that traditional microseismic source prediction methods have great limitations for complex geological conditions and harsh working environments, and the installation and maintenance costs of multiple sensors are relatively high.

[0005] To solve the above technical problems, the embodiments of this specification are implemented as follows:

[0006] In a first aspect, a microseismic source prediction method based on a single sensor provided by an embodiment of this specification includes:

[0007] Obtain microseismic signal feature data collected by a single sensor;

[0008] Input the microseismic signal feature data into a microseismic source prediction model to obtain source location information; the microseismic source prediction model is trained according to a data set of microseismic signal feature data collected by multiple single sensors.

[0009] In a second aspect, a microseismic source prediction device based on a single sensor provided by an embodiment of this specification includes:

[0010] An acquisition module, configured to obtain microseismic signal feature data collected by a single sensor;

[0011] A determination module, configured to input the microseismic signal feature data into a microseismic source prediction model to obtain source location information; the microseismic source prediction model is trained according to a data set of microseismic signal feature data collected by multiple single sensors.

[0012] In a third aspect, a microseismic source prediction device based on a single sensor provided by an embodiment of this specification includes a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to implement the microseismic source prediction method based on a single sensor in Solution 1.

[0013] In a fourth aspect, a computer-readable storage medium provided by an embodiment of this specification has a computer program stored thereon. When the computer program is executed by a processor, it implements the microseismic source prediction method based on a single sensor in Solution 1.

[0014] An embodiment of this specification achieves the following beneficial effects: By inputting the microseismic signal feature data collected by a single sensor into a microseismic source prediction model, the source location information can be obtained. Through the microseismic source prediction model trained according to a dataset of microseismic signal feature data collected by multiple single sensors, the source location can be obtained, overcoming the limitations of traditional methods that rely on accurate wave velocity information and high-precision arrival time picking. Moreover, with only one sensor, the single sensor can predict the source location, reducing the complexity of microseismic source prediction, expanding the application scenarios, being applicable to complex rock mass environments, and greatly reducing the installation and maintenance costs. Description of the Drawings

[0015] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in this application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a flowchart of a microseismic source prediction method based on a single sensor provided by an embodiment of this specification;

[0017] Figure 2 It is another flowchart of a microseismic source prediction method based on a single sensor provided by an embodiment of this specification;

[0018] Figure 3 It is a schematic diagram of the original data distribution and the data distribution after Z-Score standardization provided by an embodiment of this specification;

[0019] Figure 4 It is a schematic diagram of the working principle of the SVR model provided by an embodiment of this specification;

[0020] Figure 5 It is a schematic diagram of the visualization of the hyperparameter tuning of the SVR model provided by an embodiment of this specification;

[0021] Figure 6 Schematic diagram of the visualization of hyperparameter tuning for the GBR model provided by the embodiments of this specification;

[0022] Figure 7 Schematic diagram of the working principle and parameter optimization process of the GBR model provided by the embodiments of this specification;

[0023] Figure 8 Schematic diagram of the structure of a microseismic source prediction device based on a single sensor provided by the embodiments of this specification;

[0024] Figure 9 Schematic diagram of the structure of a microseismic source prediction device based on a single sensor provided by the embodiments of this specification. Detailed implementation manners

[0025] To make the objectives, technical solutions, and advantages of one or more embodiments of this specification clearer, the technical solutions of one or more embodiments of this specification will be clearly and completely described below in conjunction with the specific embodiments and corresponding drawings of this specification. Obviously, the described embodiments are only a part of the embodiments of this specification, rather than all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope protected by one or more embodiments of this specification.

[0026] The technical solutions provided by each embodiment of this specification will be described in detail below in conjunction with the drawings.

[0027] A microseismic source prediction method based on a single sensor provided by the embodiments of the specification will be specifically described in conjunction with the drawings.

[0028] Figure 1 Schematic flowchart of a microseismic source prediction method based on a single sensor provided by the embodiments of this specification. From a program perspective, the execution subject of the process can be a program or an application client running on an application server. On the other hand, from a hardware perspective, the execution subject of the process can be a terminal device, and this embodiment does not make special limitations on this.

[0029] As Figure 1 shown, this process may include the following steps:

[0030] Step 110: Obtain the microseismic signal feature data collected by a single sensor.

[0031] In the embodiments of this specification, sensors are arranged in a preset area on the rock surface, and the position information of the sensors is recorded. A single sensor with high sensitivity and wide bandwidth is used. The single sensor can refer to one sensor, and all the microseismic signal feature data collected by one sensor are taken as a group. The microseismic signal feature data can include energy, absolute energy, count, rise time, duration, amplitude, center frequency, peak value, etc.

[0032] Taking the spatial position where the rock to be predicted is located, a spatial rectangular coordinate system is established. This spatial rectangular coordinate system includes three coordinate axes, namely the X-axis, the Y-axis, and the Z-axis. The X-axis and the Y-axis are parallel to the horizontal plane and perpendicular to each other, while the Z-axis is perpendicular to the horizontal plane. The three coordinate axes jointly determine the positions of points in space, which are used for subsequent source location-related analysis and calculation. Before collecting microseismic signals, the coordinate origin needs to be artificially determined according to specific engineering or experimental requirements.

[0033] Step 120: Input the microseismic signal feature data into the microseismic source prediction model to obtain the source position information; the microseismic source prediction model is trained according to a dataset of microseismic signal feature data collected by multiple single sensors.

[0034] In the embodiments of this specification, the microseismic source prediction model is pre-trained based on a dataset composed of microseismic signal feature data collected by multiple single sensors.

[0035] Input a set of microseismic signal feature data to be located collected by a single sensor into the trained microseismic source prediction model. The microseismic source prediction model can output the source position information, and the position information can be the coordinates in the spatial rectangular coordinate system. The microseismic source prediction model can comprehensively analyze the relationship between microseismic signal features and the source position.

[0036] It should be understood that the order of some steps in the method described in one or more embodiments of this specification can be mutually exchanged according to actual needs, or some of the steps can also be omitted or deleted.

[0037] In the embodiments of this specification, by using the microseismic signal feature data collected by a single sensor and inputting it into the microseismic source prediction model to obtain the source position information, and through the microseismic source prediction model trained according to a dataset of microseismic signal feature data collected by multiple single sensors, the source position is obtained, which overcomes the limitations of traditional methods that rely on accurate wave velocity information and high-precision arrival time picking. Moreover, the single sensor can predict the source position only by using one sensor, which reduces the complexity of microseismic source prediction, expands the application scenarios, is applicable to complex rock mass environments, and greatly reduces the installation and maintenance costs.

[0038] Based on Figure 1For the method in [the relevant context], the embodiments of this specification also provide some specific implementation schemes of this method, which are described below.

[0039] In traditional microseismic source prediction methods, multiple sensors usually need to work together. These sensors need to be precisely arranged and calibrated, consuming a large amount of manpower. Moreover, arranging multiple sensors requires meeting the requirements of geological conditions and environmental conditions. For application scenarios where multiple sensors cannot be arranged, the prediction accuracy is greatly reduced.

[0040] To solve the above problems, optionally, the microseismic source prediction model described in the embodiments of this specification includes an SVR model and a GBR model. Inputting the microseismic signal feature data into the microseismic source prediction model to obtain the source location information may specifically include:

[0041] Input the microseismic signal feature data into the SVR model to obtain first feature data;

[0042] Fuse the first feature data with the microseismic signal feature data to obtain second feature data;

[0043] Input the second feature data into the GBR model to obtain the source location information.

[0044] In the embodiments of this specification, the microseismic source prediction model may include an SVR model and a GBR model. Input a set of microseismic signal feature data into the SVR model. The processing process of the SVR model is to map the feature data to a high-dimensional space through a kernel function (such as a linear kernel function, a Gaussian kernel function) in order to find a linear relationship in the high-dimensional space.

[0045] The SVR model fits the data by constructing a hyperplane and allows a certain error range (deviation interval band). The SVR model outputs the first feature data, and the first feature data can represent the initial position information of the source.

[0046] Fuse the first feature data output after SVR processing with the initial microseismic signal feature data collected by a single sensor. The fusion can be splicing. For example, if the initial microseismic signal feature data has 5 features and the first feature data output by the SVR model has 1 new feature, then the second feature data contains 6 features.

[0047] Input the second feature data into the GBR model. The processing process of the GBR model is to initialize a simple model (such as a constant model), and through the forward stagewise algorithm, gradually add weak learners (such as decision trees). Each weak learner fits the residual of the current model. Through grid search and cross-validation, determine the optimal parameter combination (such as the number of base learners, the maximum depth of the decision tree, the learning rate, etc.). According to the second feature data, output the source location information, such as the source location coordinates (X, Y, Z).

[0048] Optionally, before inputting the microseismic signal feature data into the microseismic source prediction model, in order for the microseismic source prediction model to learn and process data more fairly and effectively, the method may further include:

[0049] Performing normalization processing on the microseismic signal feature data to obtain processed microseismic signal feature data.

[0050] In the embodiments of this specification, different features may have different dimensions (units) and value ranges. For example, the value range of energy may be 0–10,000, while the value range of rise time may be 0–1 second. Normalization processing can eliminate the dimension differences, enabling all features to have the same scale. Normalization can be Min-Max normalization, Z-Score normalization, etc. The normalized data can prevent certain features from dominating the learning process of the model due to their large value ranges, thereby improving the overall performance of the model.

[0051] Optionally, before inputting the microseismic signal feature data into the microseismic source prediction model, in order to improve the performance, efficiency, and stability of the model, the method may further include:

[0052] Obtaining microseismic signal feature data collected by multiple single sensors and constructing an original feature dataset;

[0053] Training the microseismic source prediction model based on the original feature dataset to obtain a trained microseismic source prediction model.

[0054] In the embodiments of this specification, an experimental platform is built, multiple single sensors are arranged in the target area (such as laboratory rock samples), and a spatial rectangular coordinate system is established. The microseismic signal feature data collected by each sensor forms a group. At the same time, the position coordinates of each sensor and the true source position coordinates are obtained, and the microseismic signal feature data collected by all sensors are combined into an original feature dataset.

[0055] Before training the microseismic source prediction model, in order to eliminate the dimension differences and value range differences between different features, the original feature dataset can be normalized. The original feature dataset can be a matrix, where each row represents a sample and each column represents a feature parameter.

[0056] Taking Z-Score normalization as an example, the Z-Score normalization formula is:

[0057] where, X iIt is the microseismic signal feature data of the i-th sample, n is the number of samples in the original feature dataset, E is the average value of the feature data in the original feature dataset, and σ is the standard deviation of the feature data in the original feature dataset.

[0058] Further, optionally, in the embodiments of this specification, training the microseismic source prediction model based on the original feature dataset to obtain a trained microseismic source prediction model may specifically include:

[0059] Training the SVR model based on the original feature dataset to obtain a target SVR model;

[0060] Inputting the original feature dataset into the target SVR model to obtain a first feature dataset;

[0061] Fusing the first feature dataset with the original feature dataset to obtain a second feature dataset;

[0062] Training the GBR model based on the second feature dataset to obtain a target GBR model;

[0063] Constructing the microseismic source prediction model according to the target SVR model and the target GBR model.

[0064] In the embodiments of this specification, the SVR model is optimized. The kernel function types are set as the linear kernel function (linear) and the Gaussian kernel function (rbf), and the data is mapped to a high-dimensional space. Using grid search and cross-validation to optimize the hyperparameters of the SVR model, including: kernel function type (such as linear kernel, Gaussian kernel). The regularization parameter C takes values of 0.1, 1, and 10. Through grid search, different combinations of the kernel function and the regularization parameter are made, such as (linear, 0.1), (linear, 1), and then using cross-validation, the original feature dataset is divided into several subsets. Each time, a part of the subsets is used to train the SVR model, and another part of the subsets is used to verify the performance of the SVR model. The mean squared error (MSE) is used as the evaluation index. After calculating the MSE under different combinations, it is found that when the kernel function is the Gaussian kernel function and C = 1, the MSE is the smallest. At this time, this combination is determined as the optimal hyperparameter combination for the SVR model to process. Using this combination to process the original feature dataset, the fitting result is output as a new feature.

[0065] Training the SVR model using the optimal hyperparameter combination to obtain a target SVR model. Inputting the original feature dataset into the target SVR model, a first feature dataset and a preliminary prediction result can be obtained.

[0066] The feature data in the first feature dataset and the feature data in the original feature dataset are concatenated according to the sample correspondence. For example, the original feature dataset has 100 samples, and each sample contains 8 parameters such as energy and absolute energy, forming a 100×8 matrix. The first feature dataset has 100 samples, and each sample contains 2 new features, forming a 100×2 matrix. When concatenating, the two matrices are merged by column, and finally a 100×10 matrix is obtained, which is the second feature dataset.

[0067] Optimize the GBR model. For the parameters of the GBR model, the number of base learners n_estimators is set to 100, 200, and 300 respectively, the maximum number of layers max_depth that the decision tree is allowed to grow is set to 3, 4, and 5, and the learning rate learning_rate is set to 0.05, 0.1, and 0.2. Also use grid search and cross-validation to traverse different combinations of these parameters, such as (100, 3, 0.05), (200, 4, 0.1), etc. With the goal of minimizing the mean squared error of the prediction, it is found through calculation that when n_estimators = 200, max_depth = 4, and learning_rate = 0.05, the GBR model performs best on the validation set, and at this time the optimal parameter combination of the GBR model is obtained.

[0068] Use the GBR model with the optimal parameter combination as the target GBR model, and input the first feature dataset into the target GBR model to obtain the final prediction result.

[0069] Combine the target SVR model and the target GBR model to obtain a microseismic source prediction model. Compare the final prediction result with the true source location coordinates to effectively verify the accuracy of the microseismic source prediction model.

[0070] Specifically, input the original feature dataset into the SVR model. The goal of the SVR model is to find a suitable hyperplane f(x) = ω T φ(x i ) + b, where φ(x i ) is a non-linear mapping function that maps data from the original space to a high-dimensional feature space. In the original space, the data may not be well separated or fitted by a linear model. Through this non-linear mapping, a more suitable linear relationship can be found in the high-dimensional space; T is the transpose of the matrix; ω is the hyperplane weight vector, which determines the direction of the hyperplane; b is the bias, which is used to adjust the position of the hyperplane, and a deviation interval width of 2ε is constructed, allowing a certain error between the data points and the hyperplane.

[0071] Optimize the kernel function type (set as the linear kernel function linear and the Gaussian kernel function rbf) and the regularization parameter C (taking values 0.1, 1, 10) through grid search and cross-validation to determine the optimal hyperparameter combination, and output the SVR fitting result as a new feature.

[0072] The parameter estimation problem becomes an optimization problem, and the optimization problem formula and constraints are as follows:

[0073]

[0074] Consider introducing slack variables ξ i and to handle data points that may not satisfy the constraint conditions. The above parameter estimation problem then becomes the following optimization problem:

[0075]

[0076] Among them, ω is the weight vector of the hyperplane, which determines the direction of the hyperplane; b is the bias, used to adjust the position of the hyperplane; C is the regularization parameter, used to balance the model complexity and loss, and control the penalty degree for data points that do not satisfy the constraint conditions. A larger C value will make the model more inclined to fit the training data, but it is also prone to overfitting. A smaller C value will focus more on finding a hyperplane with a larger margin; l is the number of samples; y i is the true value of the i-th sample; f(x i ) is the predicted value of the SVR model for the i-th sample x i ; ε is the half-width of the bias interval band.

[0077] From the construction principle of the hyperplane and the interval band, the SVR model is committed to finding a suitable hyperplane f(x) = ω T φ(x i ). Taking the dataset with irregular distribution in the two-dimensional space as an example, since it is difficult to achieve effective fitting or separation through a linear model in the original two-dimensional space, the nonlinear mapping function φ(x i ) is used to map the data to a higher-dimensional space, such as a three-dimensional space. In the high-dimensional space, the direction of the hyperplane is determined by the weight vector ω, and the bias b is used to adjust the position of the hyperplane. On this basis, a bias interval band with a width of 2ε is constructed. This interval band can be regarded as the tolerance area on both sides of the hyperplane, allowing a certain degree of error between the data points within the interval band (tube) and the target function. As long as the data points are within this interval band, even if there is a deviation from the hyperplane, it is within the acceptable range.

[0078] The output results of the SVR model mainly include two parts. One is the preliminary prediction result, which will be output as a new feature. For example, a set of microseismic signal feature parameter data is taken as a sample. After being processed by the SVR model, each sample will generate a new numerical feature, which can provide richer information for the subsequent prediction of the GBR model and thus enhance its prediction effect. The other is the optimal hyperparameter combination, which mainly includes the kernel function and the regularization parameter. Through a series of calculations and comparisons, if it is determined that when the Gaussian kernel function is used as the kernel function and the regularization parameter C takes the value of 1, the model shows the best performance on the validation set, then "Gaussian kernel function, C = 1" is recognized as the optimal hyperparameter combination output.

[0079] According to the characteristics of the data, use the non - linear mapping function φ(x i ) to map the data from the original space to a high - dimensional space, and then combine the weight vector ω and the bias b to determine the hyperplane and construct the corresponding margin band. The problem of finding a suitable hyperplane and dealing with the relationship between data points and the hyperplane is transformed into an optimization problem. Suppose there are sample data (x1, y1), (x2, y2), …, (x n , y n ), and its optimization goal is to minimize the objective function where ξ i and are slack variables, and C is the regularization parameter. The search step uses the grid search method to perform a combined search for different kernel function types (such as linear kernel function, Gaussian kernel function, etc.) and the values of the regularization parameter C. For example, if there are 2 optional types of kernel functions and C has 3 different values, then 6 different parameter combinations will be generated. The validation step divides the data set into multiple subsets through cross - validation, which are used as the training set and the validation set respectively. By calculating the evaluation index (such as mean squared error) of different parameter combinations on the validation set, the performance of the model is evaluated. The fitting step is to train the entire data set with the optimal hyperparameter combination after determining it, so that the model can better fit the data, thereby outputting new features and determining the final optimal hyperparameter combination.

[0080] During the processing of the GBR model, for the number of base learners n_estimators, it is set to 100, 200, and 300 respectively for parameter optimization; the maximum number of layers max_depth that the decision tree is allowed to grow is set to 3, 4, and 5; the learning rate learning_rate is set to 0.05, 0.1, and 0.2. Through the grid search and cross - validation methods, traverse these different parameter combinations to find the parameter combination that makes the model performance reach the optimal.

[0081] The GBR model adopts the additive model, and the calculation formula of the additive model is:

[0082]

[0083] At the same time, the forward stepwise algorithm is adopted, and the calculation formula of the forward stepwise algorithm is:

[0084] f m (x) = f m-1 (x) + β m b(x; γ m )

[0085] where b(x; γ m ) is the basis function, β m is the optimal parameter learned in the function, and γ m is the parameter in the basis function. Through continuous iterative optimization, combining multiple weak classifiers to gradually reduce the prediction error, the optimal prediction result for accurately predicting the seismic source location and the best parameter set of the model are finally obtained, so as to achieve high-precision microseismic source location.

[0086] Optionally, before training the microseismic source prediction model based on the original feature dataset in the embodiments of this specification, the method may further include:

[0087] Performing normalization processing on the original feature dataset to obtain a processed original feature dataset.

[0088] In the embodiments of this specification, the original feature dataset contains multiple microseismic signal feature data. The normalization can be Min-Max normalization, Z-Score normalization, etc. Normalizing the feature data in the original feature dataset eliminates the dimension difference. The normalized data can avoid some features from dominating the learning process of the model due to a large numerical range, thereby improving the overall performance of the model.

[0089] Saving the normalized data as a new original feature dataset for subsequent model training and prediction.

[0090] Figure 2 This is another process schematic diagram of a microseismic source prediction method based on a single sensor provided by the embodiments of this specification.

[0091] As Figure 2 shown, step 210: Obtain microseismic signal feature data collected by multiple single sensors and construct an original feature dataset;

[0092] Step 220: Perform normalization processing on the original feature dataset to obtain a processed original feature dataset;

[0093] Step 230: Train an SVR model based on the processed original feature dataset to obtain a target SVR model;

[0094] Step 240: Input the original feature dataset after normalization into the target SVR model to obtain the first feature dataset;

[0095] Step 250: Fuse the first feature dataset with the original feature dataset after normalization to obtain the second feature dataset;

[0096] Step 260: Based on the second feature dataset, train the GBR model to obtain the target GBR model;

[0097] Step 270: Input the microseismic signal feature data to be located collected by a single sensor into the microseismic source prediction model integrating the target SVR model and the target GBR model to obtain the source coordinates.

[0098] In the embodiments of this specification, by using the signal feature parameters received by a single microseismic sensor and based on the microseismic source prediction model, the accuracy problem caused by the uncertainty of wave velocity in the traditional microseismic source prediction method is overcome, and the positioning accuracy is effectively improved.

[0099] The microseismic source prediction model learns and analyzes the rich signal features of a single sensor, and no longer simply relies on accurate arrival time picking for positioning. The microseismic source prediction model comprehensively considers the complex relationships between multiple microseismic parameter features, and can still accurately identify and locate the microseismic source under relatively low arrival time picking accuracy, thus effectively reducing the dependence on the arrival time picking accuracy and enhancing the reliability and stability of the entire positioning method in practical applications.

[0100] The application scenario is not limited by the number of installed sensors, and can adapt to complex engineering environments. Even when the number of sensors is limited or only one sensor can be used, microseismic source positioning can still be achieved, greatly reducing the hardware cost of the monitoring system. At the same time, the amount of data processing is reduced, and the data processing cost is lowered, effectively solving the problem of high costs caused by multi-sensor dependence, and providing a more flexible and reliable solution for rock engineering safety assurance and structural integrity assessment.

[0101] At the same time, this solution can more timely and accurately determine the position of the microseismic source inside the rock, thereby more effectively ensuring the safety of rock engineering, accurately assessing the integrity of the rock structure, early warning potential engineering risks, and ensuring the stable operation of rock engineering.

[0102] Figure 3 It is a schematic diagram of the original data distribution and the data distribution after Z-Score normalization provided by the embodiments of this specification.

[0103] As Figure 3As shown, the mean of the original data is 47.02 and the standard deviation is 29.60. After standardization, both are -0.00 and the standard deviation is 1.00. The microseismic signal characteristic data such as amplitude and rise time are processed into numerical values with a unified dimension, eliminating the dimension difference and numerical range difference between different characteristics.

[0104] Figure 4 It is a schematic diagram of the working principle of the SVR model provided by the embodiments of this specification.

[0105] As Figure 4 shown, the blue scatter points are the data points in the original feature dataset. The abscissa is the feature value and the ordinate is the target value; the red points are the support vectors, which are crucial for determining the hyperplane; the green dashed line is the regression line (hyperplane) for fitting the data trend; the two orange dashed lines are the upper boundary (+ε) and the lower boundary (-ε) respectively. The ε tolerance band formed by them and the regression line treats the fitting error of the data points within the band as acceptable.

[0106] Figure 5 It is a schematic diagram of the visualization of the hyperparameter tuning of the SVR model provided by the embodiments of this specification.

[0107] Figure 5 It includes a heat map of loss values and a three-dimensional surface plot of loss values, which are used to intuitively display the loss value situation of the model under different parameter settings. Among them, the horizontal axis in the heat map of loss values represents the type of kernel function, which may include various choices such as linear kernel function, polynomial kernel function, Gaussian kernel function, etc.; the vertical axis represents the value of the regularization parameter C. As Figure 5 shown, the heat map of loss values shows a series of C values of different magnitudes, such as 0.01, 0.03, etc.

[0108] Different colors in the heat map correspond to different loss value sizes. Generally, from blue to red indicates that the loss value increases from small to large. Through the color distribution, it can be quickly seen under which combination of kernel function types and regularization parameter C, the loss value of the model is lower, that is, the model performs better. For example, the bluer areas indicate smaller loss values under the corresponding parameter combinations, and the model has better performance in these areas.

[0109] In the three-dimensional surface plot of loss values, the three coordinate axes respectively correspond to the type of kernel function, the regularization parameter C, and the loss value. Among them, the type of kernel function and the regularization parameter C are the key hyperparameters affecting the model, and the loss value reflects the difference degree between the model prediction result and the true value.

[0110] Different points on the surface represent the loss values under different combinations of kernel function types and regularization parameter C. The undulation of the surface shows the trend of the loss value changing with the two hyperparameters. As Figure 5As shown, the lowest point (marked by the red dot) in the three-dimensional surface plot of the loss value represents the combination of the kernel function type and the regularization parameter C that can minimize the model loss value within the given parameter range. This combination is the optimal hyperparameter selection in this analysis.

[0111] Figure 6 It is a schematic diagram of the visualization of hyperparameter tuning for the GBR model provided in the embodiments of this specification.

[0112] Figure 6 It includes the heatmap of the mean squared error and the three-dimensional surface plot of the mean squared error, which are used to show the mean squared error of the model under different parameter settings, and assist in analyzing and selecting the optimal parameters. Among them, the horizontal axis in the heatmap of the mean squared error represents the learning rate, which is a key hyperparameter that controls the parameter update step size during the model training process. As Figure 6 shown, different values such as 0.05, 0.1, 0.15, 0.2, etc. are presented in the heatmap of the mean squared error; the vertical axis represents the number of trees, that is, the number of base learners in the gradient boosting regression (GBR) model, and a series of values from smaller to larger are shown in the heatmap of the mean squared error.

[0113] The heatmap of the mean squared error visually presents the magnitude of the mean squared error through different colors. Generally, from red to blue corresponds to the mean squared error from large to small. The areas with a redder color indicate that the mean squared error of the model is larger under the corresponding combination of the number of trees and the learning rate, which means that the difference between the model prediction value and the true value is larger and the model performance is poor; while the areas with a bluer color indicate that the mean squared error is smaller and the prediction effect of the GBR model under this parameter combination is better. The "threshold 15" in the heatmap of the mean squared error is a reference value used to assist in judging whether the model performance meets the standard.

[0114] In the three-dimensional surface plot of the mean squared error, the three coordinate axes are respectively the number of trees, the learning rate, and the mean squared error. The number of trees and the learning rate are hyperparameters that can be adjusted during the model training process, and the mean squared error is used to measure the accuracy of the model prediction.

[0115] Each point on the surface corresponds to the mean squared error value under a specific combination of the number of trees and the learning rate. The undulation of the surface reflects the trend of the mean squared error changing with these two hyperparameters. The red dot in the three-dimensional surface plot of the mean squared error marks the lowest point within the given parameter range. The combination of the number of trees and the learning rate corresponding to this point can minimize the mean squared error, which represents the optimal hyperparameter selection in this analysis and helps to improve the prediction performance of the model.

[0116] Figure 7 It is a schematic diagram of the working principle and parameter optimization process of the GBR model provided in the embodiments of this specification.

[0117] As Figure 7As shown, during the iterative optimization process of the GBR model, the output values of the weak learners show a significant decreasing trend as the number of iterations increases. This dynamic change intuitively presents the parameter adjustment and performance optimization process of the model under the iterative optimization mechanism. For the parameter optimization part, the research designs multiple groups of parameter combination experiments (including core parameters such as n_estimators, max_depth, and learning_rate), systematically compares the Performance index values corresponding to different parameter configurations, and thus explores the influence law of parameters on the model performance, providing data support for the optimal configuration of the GBR model parameters. Further analyzing the change trend of the error with the number of iterations, it can be seen that during the iterative learning process, the model error shows a continuous convergence feature. This trend fully indicates that the GBR model continuously adjusts its internal parameters and structure through iterative optimization to achieve a progressive reduction in the prediction error, ultimately promoting the performance of the GBR model to evolve towards a better state.

[0118] Figure 8 FIG. is a schematic structural diagram of a microseismic source prediction device based on a single sensor provided by an embodiment of the present specification.

[0119] Corresponding to the method embodiment, this embodiment further provides a microseismic source prediction device based on a single sensor, which may include:

[0120] An acquisition module 802, configured to acquire microseismic signal feature data collected by a single sensor;

[0121] A determination module 804, configured to input the microseismic signal feature data into a microseismic source prediction model to obtain source location information; the microseismic source prediction model is trained according to a dataset of microseismic signal feature data collected by multiple single sensors.

[0122] Optionally, in the embodiment of the present specification, the microseismic source prediction model includes an SVR model and a GBR model. The inputting the microseismic signal feature data into the microseismic source prediction model to obtain source location information may specifically include:

[0123] Inputting the microseismic signal feature data into the SVR model to obtain first feature data;

[0124] Fusing the first feature data with the microseismic signal feature data to obtain second feature data;

[0125] Inputting the second feature data into the GBR model to obtain source location information.

[0126] Based on the same idea, this embodiment of the present specification further provides a device corresponding to the above method.

[0127] Figure 9This is a schematic structural diagram of a microseismic source prediction device based on a single sensor provided by an embodiment of this specification. As Figure 9 shown, a microseismic source prediction device 900 based on a single sensor provided by an embodiment of this specification includes a memory 930, a processor 910, and a computer program 920 stored in the memory. The processor 910 executes the computer program 920 to implement the microseismic source prediction method based on a single sensor described in any of the above embodiments.

[0128] A microseismic source prediction device based on a single sensor provided by an embodiment of this specification may include a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the microseismic source prediction method based on a single sensor described in any of the above embodiments.

[0129] A computer-readable storage medium provided by an embodiment of this specification stores a computer program, and when the computer program is executed by a processor, it can implement the microseismic source prediction method based on a single sensor described in any of the above embodiments.

[0130] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments. In particular, for Figure 9 the device shown, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant parts, reference can be made to the partial description of the method embodiment.

[0131] In the 1990s, it was possible to clearly distinguish whether an improvement to a technology was a hardware improvement (e.g., improvement to circuit structures such as diodes, transistors, switches, etc.) or a software improvement (improvement to method flows). However, with the development of technology, many method flow improvements today can be regarded as direct improvements to hardware circuit structures. Almost all designers obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement to a method flow cannot be implemented using a hardware entity module. For example, a Programmable Logic Device (PLD) (e.g., a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logical function is determined by the user programming the device. Designers can program on their own to "integrate" a digital system onto a single PLD, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly implemented using "logic compiler" software, which is similar to the software compiler used in program development and writing. The original code before compilation also has to be written in a specific programming language, which is called a Hardware Description Language (HDL). There is not just one type of HDL, but many types, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. Currently, the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also be aware that by simply performing some logical programming on the method flow using the above-mentioned several hardware description languages and programming it into the integrated circuit, it is easy to obtain the hardware circuit that implements the logical method flow.

[0132] The controller can be implemented in any suitable manner. For example, the controller can take the form of, for example, a microprocessor or a processor and a computer-readable medium storing computer-readable program code (such as software or firmware) executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of the controller include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that, in addition to implementing the controller in the form of pure computer-readable program code, it is entirely possible to make the controller implement the same function in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the devices included therein for implementing various functions can also be regarded as the structures within the hardware component. Or even, the devices for implementing various functions can be regarded as either software modules for implementing the method or the structures within the hardware component.

[0133] The systems, devices, modules, or units illustrated in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or any combination of these devices.

[0134] For the convenience of description, when describing the above devices, they are described separately as various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0135] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0136] The present invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each flow and / or block in the flowchart illustrations and / or block diagrams, and combinations of flows and / or blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing apparatus create means for implementing the functions specified in the flowchart Figure 1 for one or more flows and / or blocks Figure 1 for one or more blocks.

[0137] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means that implement the functions specified in the flowchart Figure 1 for one or more flows and / or blocks Figure 1 for one or more blocks.

[0138] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart Figure 1 for one or more flows and / or blocks Figure 1 for one or more blocks.

[0139] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0140] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer-readable media.

[0141] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0142] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0143] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0144] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0145] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A microseismic source prediction method based on a single sensor, characterized in that: include: Obtain characteristic data of microseismic signals collected by a single sensor; Inputting the microseismic signal characteristic data into a microseismic source prediction model to obtain source location information; The microseismic source prediction model is trained based on a data set of microseismic signal characteristic data collected by multiple single sensors.

2. The method according to claim 1, characterized in that The microseismic source prediction model includes an SVR model and a GBR model, and the step of inputting the microseismic signal characteristic data into the microseismic source prediction model to obtain the source location information specifically includes: Inputting the microseismic signal characteristic data into the SVR model to obtain first characteristic data; fusing the first characteristic data with the microseismic signal characteristic data to obtain second characteristic data; The second characteristic data is input into the GBR model to obtain the earthquake source location information.

3. The method according to claim 1, characterized in that Before inputting the microseismic signal characteristic data into the microseismic source prediction model, the method further includes: The microseismic signal characteristic data is standardized to obtain processed microseismic signal characteristic data.

4. The method according to claim 1, characterized in that: Before inputting the microseismic signal characteristic data into the microseismic source prediction model, the method further includes: Acquire characteristic data of microseismic signals collected by multiple single sensors and construct an original characteristic data set; The microseismic source prediction model is trained based on the original feature data set to obtain a trained microseismic source prediction model.

5. The method according to claim 4, characterized in that The step of training the microseismic source prediction model based on the original feature data set to obtain a trained microseismic source prediction model specifically includes: Based on the original feature data set, the SVR model is trained to obtain a target SVR model; Inputting the original feature data set into the target SVR model to obtain a first feature data set; Fusion the first feature data set with the original feature data set to obtain a second feature data set; Based on the second feature number set, the GBR model is trained to obtain a target GBR model; The microseismic source prediction model is constructed according to the target SVR model and the target GBR model.

6. The method according to claim 4, characterized in that Before training the microseismic source prediction model based on the original feature data set, the method further includes: The original feature data set is standardized to obtain a processed original feature data set.

7. A microseismic source prediction device based on a single sensor, characterized in that: include: An acquisition module is used to acquire characteristic data of microseismic signals collected by a single sensor; The determination module is used to input the microseismic signal characteristic data into a microseismic source prediction model to obtain source location information; the microseismic source prediction model is trained based on a data set of microseismic signal characteristic data collected by multiple single sensors.

8. The device according to claim 7, characterized in that The microseismic source prediction model includes an SVR model and a GBR model, and the step of inputting the microseismic signal characteristic data into the microseismic source prediction model to obtain the source location information specifically includes: Inputting the microseismic signal characteristic data into the SVR model to obtain first characteristic data; fusing the first characteristic data with the microseismic signal characteristic data to obtain second characteristic data; The second characteristic data is input into the GBR model to obtain the earthquake source location information.

9. A microseismic source prediction device based on a single sensor, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images