Earthquake phase identification method and system based on classification algorithm

Abstract

The invention relates to the technical field of data processing, in particular to a seismic phase identification method and system based on a classification algorithm, and the method comprises the steps: obtaining original seismic wave data, generating standard seismic wave data according to the waveform characteristics of the seismic wave data, and forming a feature vector set; extracting multi-dimensional classification features from the standard seismic wave data in the feature vector set, and constructing the feature vector set based on the multi-dimensional classification features; training a decision tree through the feature vector set to obtain a trained decision tree; training a naive Bayes classifier through the feature vector set to obtain a trained classifier; performing weighted fusion on the trained decision tree and the trained classifier to obtain a fusion model; performing seismic phase identification on the seismic wave data based on the fusion model, and outputting an identification result of a standard seismic phase and an abnormal seismic phase; according to the invention, the accuracy and efficiency of seismic phase recognition are improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a seismic phase identification method and system based on a classification algorithm. Background Technology

[0002] Traditional earthquake phase identification methods often rely on human experience and simple threshold judgments, which are inefficient and have limited accuracy. With the continuous increase in the volume and complexity of earthquake monitoring data, traditional methods struggle to meet the demands for high-precision phase identification. Furthermore, seismic wave signals are subject to various interferences during propagation, such as geological structures and noise, further increasing the difficulty of phase identification. Therefore, researching classification-based phase identification methods and systems is of significant practical importance for more accurate and efficient phase identification, enabling more reliable earthquake early warning and analysis. Summary of the Invention

[0003] The purpose of this invention is to provide a seismic phase identification method and system based on a classification algorithm, aiming to improve the accuracy and efficiency of seismic phase identification.

[0004] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a seismic phase identification method based on a classification algorithm, the method comprising the following steps: S100: Acquire raw seismic wave data, generate standard seismic wave data based on the waveform characteristics of the seismic wave data, label the standard and anomalous phases in the standard seismic wave data, and form a set of feature vectors. S200 extracts multi-dimensional classification features from standard seismic wave data in the feature vector set, and constructs a feature vector set based on the multi-dimensional classification features; S300 trains a decision tree using a set of feature vectors, and uses the decision tree to mine the non-linear association rules between the features in the set of feature vectors, thus obtaining a trained decision tree. S400 trains a Naive Bayes classifier using a set of feature vectors, calculates the global probability of seismic phase types using the Naive Bayes classifier, and obtains a trained classifier. S500 performs weighted fusion of the trained decision tree and the trained classifier to obtain a fusion model; based on the fusion model, it performs phase identification on seismic wave data and outputs the identification results of standard phase and anomalous phase.

[0005] Optionally, in S100, the process of acquiring raw seismic wave data, generating standard seismic wave data based on the waveform characteristics of the seismic wave data, labeling the standard and anomalous phases in the standard seismic wave data, and forming a feature vector set includes: S110: Obtain the raw seismic wave data, filter the seismic wave data according to the waveform characteristics, and obtain the filtered seismic wave data. S120 converts the filtered seismic wave data into a unified standard format to obtain standard seismic wave data. S130, based on the waveform characteristics of standard seismic wave data, label the standard seismic phase and anomalous seismic phase in the standard seismic wave data; S140 forms a complete set of feature vectors from the labeled standard seismic wave data.

[0006] Optionally, in S200, the step of extracting multi-dimensional classification features from standard seismic wave data in the feature vector set and constructing a feature vector set based on the multi-dimensional classification features includes: Standard seismic wave data is converted into a structured feature matrix. The feature matrix is ​​then discretized to obtain multiple discrete features, which serve as a set of feature vectors. The feature matrix contains both dynamic and statistical features.

[0007] Optionally, in S300, training the decision tree using a set of feature vectors and using the decision tree to mine non-linear association rules between features in the set of feature vectors to obtain a trained decision tree includes: S310, create an empty decision tree. The nodes of the decision tree represent features, the branches represent the values ​​of the features, the leaf nodes are used to represent the classification results, and the classification results represent the seismic phase type. S320, calculate the information gain ratio of each feature, select the feature with the largest information gain ratio as the splitting attribute of the current node, and generate multiple child nodes for the selected splitting attribute according to the discretization threshold. S330: Determine whether the generated child nodes meet the termination condition. If yes, execute S330; otherwise, execute S320. The termination condition is that the number of child node samples is less than the set quantity threshold or the purity reaches the set proportion threshold. S340 employs a pessimistic error pruning strategy to prune the constructed decision tree, forming classification rules from the root node to the leaf node, thus obtaining a trained decision tree.

[0008] Optionally, in S400, the step of training a Naive Bayes classifier using a set of feature vectors and calculating the global probability of the seismic phase type using the Naive Bayes classifier to obtain a trained classifier includes: S410, perform data preprocessing on the feature vector set, dividing the feature vector set into a training set and a test set; S420: Based on the frequency of occurrence of features of various earthquake phases in the training set, calculate the conditional probability and prior probability of each feature under different earthquake phase categories. According to Bayes' theorem, multiply the prior probability and the conditional probability of each feature to obtain the posterior probability of each earthquake phase type. Based on the posterior probability of each earthquake phase type, obtain the global probability of the earthquake phase type. S430 uses the test set to evaluate the performance of the trained Naive Bayes classifier until the evaluation criteria are met, thus obtaining the trained classifier.

[0009] Optionally, in S500, the step of identifying seismic wave data based on the fusion model and outputting the identification results of standard and anomalous seismic phases includes: S510 inputs newly acquired seismic wave data into the fusion model to obtain the class probability output by the decision tree and the posterior probability output by the Naive Bayes. S520 performs a weighted fusion of the class probability and the posterior probability to obtain the fused probability; the class with the highest probability is selected as the final classification result.

[0010] Optionally, after S500, the method further includes: Set a confidence threshold for anomalous seismic phases. When the classification results of the seismic wave data by the trained decision tree and the trained classifier are consistent, and the weighted probability reaches the probability threshold, the seismic wave data is marked as a high-confidence anomalous signal.

[0011] Secondly, embodiments of the present invention provide a seismic phase identification system based on a classification algorithm, the system comprising: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor performs the method as described in any of the preceding statements.

[0012] The beneficial effects of this invention are as follows: By acquiring raw seismic wave data and generating standard seismic wave data, constructing a feature vector set, and training decision trees and Naive Bayes classifiers separately before weighted fusion, this invention can fully explore the association rules between data features, comprehensively utilize the advantages of different algorithms, and effectively improve the accuracy and efficiency of seismic phase identification. Compared with traditional methods that rely on human experience and simple threshold judgments, this method can better cope with increasingly complex earthquake monitoring data and interference factors in the signal propagation process, providing a more reliable basis for earthquake early warning and analysis. This invention can quickly and accurately identify standard seismic phases and anomalous seismic phases. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 This is a flowchart illustrating the seismic phase identification method based on a classification algorithm in an embodiment of the present invention. Figure 2 This is a schematic diagram of the seismic phase identification system based on a classification algorithm in an embodiment of the present invention. Detailed Implementation

[0015] The following will clearly and completely describe the concept, specific structure, and technical effects of the present invention in conjunction with embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present invention can be combined with each other.

[0016] See Figure 1 This invention provides a seismic phase identification method based on a classification algorithm, the method comprising the following steps: S100: Acquire raw seismic wave data, generate standard seismic wave data based on the waveform characteristics of the seismic wave data, label the standard and anomalous phases in the standard seismic wave data, and form a set of feature vectors. S200 extracts multi-dimensional classification features from standard seismic wave data in the feature vector set, and constructs a feature vector set based on the multi-dimensional classification features; S300 trains a decision tree using a set of feature vectors, and uses the decision tree to mine the non-linear association rules between the features in the set of feature vectors, thus obtaining a trained decision tree. S400 trains a Naive Bayes classifier using a set of feature vectors, calculates the global probability of seismic phase types using the Naive Bayes classifier, and obtains a trained classifier. S500 performs weighted fusion of the trained decision tree and the trained classifier to obtain a fusion model; based on the fusion model, it performs phase identification on seismic wave data and outputs the identification results of standard phase and anomalous phase.

[0017] In the embodiments provided by this invention, firstly, based on the waveform characteristics of the original seismic wave data, noise interference is eliminated and the data format is unified through data preprocessing, while simultaneously labeling a set of feature vectors containing standard seismic phases (such as S, ScS) and anomalous seismic phases (such as SdS, PdP). Then, multi-dimensional classification features are extracted from the preprocessed data, including dynamic and statistical parameters such as wave crests, troughs, amplitudes, periods, cross-correlation coefficients, and relative travel time differences of the seismic phases, to construct a set of feature vectors. Secondly, adopting the principle of "dual-mode collaboration and probability weighting," decision trees (C4.5 algorithm) are used to mine nonlinear association rules between features, and a Naive Bayes classifier is used to calculate the global probability distribution of seismic phase types. The two are optimized through a weighted fusion strategy to optimize the classification boundary. Finally, the fused model is applied to actual seismic wave data, automatically outputting the identification results of standard and anomalous seismic phases, and marking high-confidence anomalous signals, completing the rapid screening and classification of seismic data, ultimately achieving efficient identification of complex seismic phases and accurate detection of deep Earth structures.

[0018] "Dual-mode collaboration with probabilistic weighting": This means using a decision tree model to analyze the hierarchical logical relationships between features (such as the coupling effect between travel time and amplitude), while using a Naive Bayes model to quantify the global distribution pattern of seismic phase categories from a probabilistic perspective (such as the prior probability of SdS occurrence). After training both models separately, the prediction results are weighted (60% for decision tree and 40% for Naive Bayes) to balance feature correlation and statistical robustness, thereby improving the model's ability to generalize and identify additional seismic phases.

[0019] In some embodiments, S100, the process of acquiring raw seismic wave data, generating standard seismic wave data based on the waveform characteristics of the seismic wave data, labeling the standard and anomalous phases in the standard seismic wave data, and forming a feature vector set includes: S110: Obtain the raw seismic wave data, filter the seismic wave data according to the waveform characteristics, and obtain the filtered seismic wave data. First, raw seismic wave data is collected, which may come from different seismic monitoring stations and different types of seismic events. Next, based on the frequency characteristics of the seismic wave data and the noise frequency range, an appropriate filter is selected. For example, a low-pass filter can be used for high-frequency noise, allowing seismic wave signals below a certain frequency to pass while blocking high-frequency noise; a high-pass filter is used for low-frequency noise. By setting a reasonable cutoff frequency, noise components are effectively removed, making the seismic wave data signal clearer.

[0020] S120 converts the filtered seismic wave data into a unified standard format to obtain standard seismic wave data. Specifically, examine the data formats of raw seismic wave data from different sources. This data may be stored in different file formats, such as text files and binary files, and the data storage structure and encoding methods may also differ. Establish a unified data format standard. This includes determining the data storage file format (such as common general-purpose seismic data formats), specifying the data arrangement order (such as time series), data types (such as integers and floating-point numbers), and data precision. Convert seismic wave data from different formats to a unified standard format. During the conversion process, ensure data integrity and accuracy, and avoid data loss or errors.

[0021] S130, based on the waveform characteristics of standard seismic wave data, label the standard seismic phase and anomalous seismic phase in the standard seismic wave data; Acquiring the waveform characteristics of seismic wave data involves analyzing parameters such as amplitude, frequency, period, and phase to understand the distinctive features of different seismic phases (standard and anomalous phases) in these parameters. For example, standard S-waves typically exhibit specific propagation velocities and waveform characteristics, with their frequency and amplitude varying regularly within a certain range; anomalous SdS waves may display unique waveform distortions or delays. By studying and comparing a large amount of historical data, these characteristic patterns can be summarized, providing a basis for subsequent processing and annotation.

[0022] Standard seismic phases (such as S-waves and ScS-waves) are identified from standard seismic wave data by utilizing their waveform characteristics. Rule-based methods can be used to determine these phases based on their propagation time, frequency range, amplitude variations, and other characteristics. For example, the theoretical propagation time of the S-wave can be calculated based on the earthquake's location and the location of the monitoring station. Waveform segments matching the frequency and amplitude characteristics of the S-wave can then be identified around this time. The identified standard seismic phases are then labeled. Labeling information typically includes the phase type (e.g., S-wave or ScS), start time, and end time. This labeling information is then correlated with the corresponding seismic wave data to form labeled data for the standard seismic phases.

[0023] Similarly, based on the waveform characteristics of anomalous seismic phases, anomalous phases (such as SdS and PdP) are identified in seismic wave data. Since anomalous phases are relatively rare and their characteristics are more complex, identification may require combining more prior knowledge and machine learning algorithms. For example, a machine learning classifier can be used to classify seismic wave data and determine whether it is an anomalous phase. Identified anomalous phases are then labeled, with annotations similar to those for standard phases, including phase type, start time, and end time. The accuracy and consistency of the annotations are crucial for subsequent model training.

[0024] S140 forms a complete set of feature vectors from the labeled standard seismic wave data.

[0025] In some embodiments, in S200, the step of extracting multi-dimensional classification features from standard seismic wave data in the feature vector set and constructing a feature vector set based on the multi-dimensional classification features includes: Standard seismic wave data is converted into a structured feature matrix. The feature matrix is ​​then discretized to obtain multiple discrete features, which serve as a set of feature vectors. The feature matrix contains both dynamic and statistical features.

[0026] Specifically, dynamic features include peak / trough amplitude, phase period, and waveform energy attenuation rate; statistical features include cross-correlation coefficients of adjacent phases, relative travel time difference, and waveform similarity index. The feature matrix is ​​discretized, and continuous features (such as amplitude and travel time difference) are binned and converted into discrete values.

[0027] In some embodiments, in S300, training the decision tree using a set of feature vectors and using the decision tree to mine non-linear association rules between features in the set of feature vectors to obtain a trained decision tree includes: S310, create an empty decision tree. The nodes of the decision tree represent features, the branches represent the values ​​of the features, the leaf nodes are used to represent the classification results, and the classification results represent the seismic phase type. S320, calculate the information gain ratio of each feature, select the feature with the largest information gain ratio as the splitting attribute of the current node, and generate multiple child nodes for the selected splitting attribute according to the discretization threshold. Specifically, the information gain ratio (IGR) of each feature is calculated when it partitions the feature vector set at the current node. The larger the IIGR, the greater the impact of the feature on classification, and the more suitable it is as the splitting feature for the current node. The formula for calculating the IIGR for each feature is: IIGR = Information Gain / Intrinsic Information. The feature with the largest IIGR is selected as the splitting attribute for the current node, as this feature has the strongest ability to distinguish different seismic phase types.

[0028] Based on the selected splitting attribute, the feature vector set is partitioned, generating nodes and branches of the decision tree. At each node, based on the value of the current best splitting feature, the feature vector set is divided into different subsets, each subset corresponding to a branch. For example, if the current best splitting feature is peak amplitude, and an amplitude threshold is set, then samples in the dataset with peak amplitudes greater than the threshold will be assigned to one branch, and samples with peak amplitudes less than the threshold will be assigned to another branch.

[0029] S330: Determine whether the generated child nodes meet the termination condition. If yes, execute S330; otherwise, execute S320. The termination condition is that the number of child node samples is less than the set quantity threshold or the purity reaches the set proportion threshold. Repeat step S230 at each child node until a termination condition is met (such as the number of samples in the child node being less than a set threshold or the purity reaching 95%). By recursively performing the above splitting process on the divided subsets until a stop condition is met. The stop conditions are: all samples in the subset belong to the same category, at which time a leaf node is created and labeled with that category; or there are no remaining features to choose from, at which time a leaf node is created and labeled with the category that appears most frequently in the subset. When the stop condition is met, the node becomes a leaf node and is labeled with the corresponding category (such as a standard seismic phase or an abnormal seismic phase). For the selected splitting attribute, generate child nodes according to the discretization threshold, and each child node corresponds to a value range of the feature. For example, the cross-correlation coefficient ≤ 0.6 is one category, and the cross-correlation coefficient > 0.6 is another category.

[0030] S340, Prune the constructed decision tree using the pessimistic error pruning strategy. The path from the root node to the leaf node forms a classification rule, and the trained decision tree is obtained.

[0031] Prune the generated decision tree to prevent overfitting. The decision tree after pruning is the trained decision tree, which can, based on the input set of feature vectors, discover the non-linear association rules between various features, thereby classifying and predicting seismic phases.

[0032] Specifically, adopt the pessimistic error pruning (PEP) strategy to remove subtrees that contribute little to classification and prevent overfitting. The path from the root node to the leaf node forms a classification rule. For example: If the travel time difference between the S and ScS seismic phases satisfies t_S < t_SdS < t_ScS and is highly similar to the S seismic phase (correlation coefficient > 0.6), then it is determined as the SdS seismic phase. Here, t_S represents the travel time of the S seismic phase, t_SdS represents the travel time of the SdS seismic phase, and t_ScS represents the travel time of the ScS seismic phase. Through the above steps, non-linear association rules between features are discovered, providing a basis for subsequent result fusion with the naive Bayes classifier.

[0033] The following is an implementation example of the decision tree analyzing the hierarchical logical relationship between features; Example: Analysis of the coupled influence of travel time difference and amplitude; Data samples: Include 600 seismic wave records, among which 150 are labeled as abnormal seismic phases (SdS).

[0034] Correction of the core features of the SdS seismic phase; Location feature: Located between the S wave and the ScS wave (t_S < t_SdS < t_ScS); Genetic feature: The reflected seismic phase caused by the abnormal body in the D" discontinuity; Waveform feature: Amplitude: typically 40-70% of the S-wave amplitude; Waveform: Highly similar to the S-wave (correlation coefficient > 0.6); Example of a decision tree splitting process:

[0035] Division process: Root node (600 samples): Splitting characteristics: t_S <t<t_ScS; Result of the split: Satisfaction window: 400 samples (including 142SdS); Not satisfied: 200 samples (including 8 SdS) are judged as non-SdS; First layer split (400 samples): Splitting characteristics: t_rel∈[0.2,0.8]; Result of the split: Satisfaction range: 350 samples (including 140SdS) (excluding the 20% regions at both ends of the window to reduce boundary effects) Not satisfied: 50 samples (including 2SdS) are identified as S-wave tail waves; Second layer split (350 samples): Splitting characteristics: A_ratio∈[0.4,0.7]; Result of the split: Satisfaction range: 300 samples (including 135SdS); The following condition was not met: 50 samples (including 5SdS) were considered noise. Third layer split (300 samples): Splitting characteristics: CC_S > 0.6; Result of the split: Satisfies: 280 samples (including 132 SdS); The following condition was not met: 20 samples (including 3SdS) were identified as instrument interference. Fourth layer split (280 samples): Splitting characteristic: CC_noise < 0.3; Result of the split: If the following condition is met: 260 samples (including 130 SdS), it is determined to be SdS; Not satisfied: 20 samples (including 2SdS) were identified as environmental noise.

[0036] In some embodiments, S400, training the Naive Bayes classifier using a set of feature vectors, calculating the global probability of the seismic phase type using the Naive Bayes classifier, and obtaining the trained classifier includes: S410, perform data preprocessing on the feature vector set, dividing the feature vector set into a training set and a test set; It should be noted that the training set is used to train the Naive Bayes classifier, and the test set is used to evaluate the classifier's performance. First, the data size and distribution of the feature vector set are analyzed. Based on the characteristics of the data, the feature vector set is randomly divided according to a certain ratio (e.g., 70% as the training set and 30% as the test set). During the partitioning process, it is important to ensure that the data distributions of the training and test sets are as similar as possible to ensure the accuracy of the evaluation results. For example, if the ratio of standard seismic phases to anomalous seismic phases in the feature vector set is 3:1, then this ratio should also be maintained as much as possible in the training and test sets.

[0037] S420: Based on the frequency of occurrence of features of various earthquake phases in the training set, calculate the conditional probability and prior probability of each feature under different earthquake phase categories. According to Bayes' theorem, multiply the prior probability and the conditional probability of each feature to obtain the posterior probability of each earthquake phase type. Based on the posterior probability of each earthquake phase type, obtain the global probability of the earthquake phase type. During training, each feature is assumed to be independent. Based on the frequency of feature occurrences across different seismic phases (standard and anomalous phases) in the training set, the conditional probability of each feature under different phase categories is calculated. For example, for a feature (such as wave crest amplitude), the probability of that amplitude value occurring under standard and anomalous phases is calculated. Taking wave crest amplitude as an example, the number of samples in the S-phase sample where the wave crest amplitude falls within a specific interval, and the number of samples in the SdS-phase sample where the wave crest amplitude falls within the same interval, are counted, and then the conditional probability of wave crest amplitude under S-phase and SdS-phase is calculated. For other features, such as phase period and waveform energy attenuation rate, the conditional probabilities under different phase types are calculated in the same way. Then, the prior probabilities of each type of seismic phase (i.e., the probability of each type of seismic phase appearing in the overall data) are combined. For example, if there are 1000 samples in the feature vector set, of which there are 600 S-phases and 150 SdS-phases, then the prior probability of the S-phase is 60%, the prior probability of the SdS-phase is 15%, and the others account for 25%.

[0038] For a given seismic wave data sample, its feature vector includes features such as peak amplitude and phase period. By multiplying the conditional probabilities of these features under different phase types with the prior probabilities of the corresponding phase types, the posterior probabilities of the sample belonging to different phase types are obtained. The phase type with the highest posterior probability is the classification result of the Naive Bayes classifier for that sample. Through statistical analysis of different phase types and features in the feature vector set, the global probability distribution of phase types can be accurately calculated.

[0039] S430 uses the test set to evaluate the performance of the trained Naive Bayes classifier until the evaluation criteria are met, thus obtaining the trained classifier.

[0040] Specifically, evaluation metrics include accuracy, recall, and F1 score. Feature vectors from the test set are input into the trained classifier to obtain classification prediction results. These results are compared with the actual seismic phase categories in the test set to calculate accuracy (the proportion of correctly predicted samples out of the total number of samples), recall (the proportion of correctly predicted samples of a certain seismic phase out of the actual number of samples of that type), and F1 score (the harmonic mean of accuracy and recall). These metrics are used to evaluate the classifier's performance in seismic phase type identification and determine whether the classifier has achieved the expected results. If the performance is unsatisfactory, training parameters can be adjusted or retraining can be performed.

[0041] After training and evaluation, a well-trained classifier is obtained. This classifier can calculate the global probability distribution of seismic phase types based on the input feature vector, providing a basis for subsequent weighted fusion with the results of the decision tree. For example, for a new seismic wave data feature vector, the well-trained classifier can output the probability that the data belongs to the standard seismic phase and the anomalous seismic phase, thus providing an important basis for comprehensively judging the seismic phase type.

[0042] In some embodiments, the prior probability formula is: Among them, C i This represents the type of the i-th seismic phase. Represents the seismic phase type in the eigenvector set. The number of samples, where N represents the total number of samples.

[0043] In some embodiments, the formula for calculating the posterior probability is: ; Where X is the input feature vector, X j For the j-th feature, Seismic phase type C i The posterior probability.

[0044] In some embodiments, S500, the step of identifying seismic phases based on a fusion model and outputting the identification results of standard and anomalous seismic phases includes: S510 inputs newly acquired seismic wave data into the fusion model to obtain the class probability output by the decision tree and the posterior probability output by the Naive Bayes. S520 performs a weighted fusion of the class probability and the posterior probability to obtain the fused probability; the class with the highest probability is selected as the final classification result.

[0045] Specifically, first, weight distribution is performed. For example, the decision tree weight is 60% (emphasizing complex associations between features), and the Naive Bayes weight is 40% (emphasizing global statistical distributions). Then, weighted voting is carried out, and the probability after fusion = 0.6 × class probability + 0.4 × posterior probability; finally, the class with the highest probability is selected as the final classification result.

[0046] In some embodiments, after S500, the method further includes: Setting a confidence threshold for abnormal seismic phases. When the classification results of the trained decision tree and the trained classifier for seismic wave data are consistent and the weighted probability reaches the probability threshold, then mark this seismic wave data as a high-confidence abnormal signal.

[0047] In the embodiments provided by the present invention, it is also possible to automatically output the recognition results of standard seismic phases and abnormal seismic phases, and mark high-confidence abnormal signals. The implementation method of the automatic marking rule for high-confidence abnormal signals is as follows: Confidence threshold setting: According to the verification results of the test set, set the confidence threshold for abnormal seismic phases (such as SdS), for example, 90%; if the probability of SdS output by the model ≥ 90%, then mark it as a "high-confidence abnormal signal".

[0048] Multi-model collaborative verification: When the classification results of the decision tree and Naive Bayes for a certain sample are consistent (both are SdS) and the weighted probability ≥ 85%, then directly mark it.

[0049] Example: Input sample features: travel time difference t_S < t_SdS < t_ScS, amplitude = 0.6, cross-correlation coefficient = 0.6; Decision tree output: SdS probability 95%; Naive Bayes output: SdS probability 88%; Weighted probability: 0.6 × 0.95 + 0.4 × 0.88 = 0.9220.6 × 0.95 + 0.4 × 0.88 = 0.922; Judgment result: High-confidence abnormal signal (marked as a red warning).

[0050] Corresponding to Figure 1 the method of Figure 2 In the embodiments of the present invention, a seismic phase recognition system based on a classification algorithm is provided, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0051] It is evident that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0052] Furthermore, embodiments of the present invention also disclose a computer program product or computer program stored in a computer-readable storage medium. A processor of a computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program, causing the computer device to perform the described method. Similarly, the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0053] It will be understood by those skilled in the art that all or some of the methods and systems disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0054] The above is a detailed description of the preferred embodiments of this disclosure. However, this disclosure is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this disclosure. All such equivalent modifications or substitutions are included within the scope defined by the claims of this disclosure.

Claims

1. A seismic phase identification method based on a classification algorithm, characterized in that, The method includes the following steps: S100: Acquire raw seismic wave data, generate standard seismic wave data based on the waveform characteristics of the seismic wave data, label the standard and anomalous phases in the standard seismic wave data, and form a set of feature vectors. S200 extracts multi-dimensional classification features from standard seismic wave data in the feature vector set, and constructs a feature vector set based on the multi-dimensional classification features; S300 trains a decision tree using a set of feature vectors, and uses the decision tree to mine the non-linear association rules between the features in the set of feature vectors, thus obtaining a trained decision tree. S400 trains a Naive Bayes classifier using a set of feature vectors, calculates the global probability of seismic phase types using the Naive Bayes classifier, and obtains a trained classifier. S500 performs weighted fusion of the trained decision tree and the trained classifier to obtain a fusion model; based on the fusion model, it performs phase identification on seismic wave data and outputs the identification results of standard phase and anomalous phase.

2. The method according to claim 1, characterized in that, In S100, the process of acquiring raw seismic wave data, generating standard seismic wave data based on the waveform characteristics of the seismic wave data, labeling the standard and anomalous phases in the standard seismic wave data, and forming a feature vector set includes: S110: Obtain the raw seismic wave data, filter the seismic wave data according to the waveform characteristics, and obtain the filtered seismic wave data. S120 converts the filtered seismic wave data into a unified standard format to obtain standard seismic wave data. S130, based on the waveform characteristics of standard seismic wave data, label the standard seismic phase and anomalous seismic phase in the standard seismic wave data; S140 forms a complete set of feature vectors from the labeled standard seismic wave data.

3. The method according to claim 1, characterized in that, In S200, the step of extracting multi-dimensional classification features from standard seismic wave data in the feature vector set and constructing a feature vector set based on the multi-dimensional classification features includes: Standard seismic wave data is converted into a structured feature matrix. The feature matrix is ​​then discretized to obtain multiple discrete features, which serve as a set of feature vectors. The feature matrix contains both dynamic and statistical features.

4. The method according to claim 1, characterized in that, In S300, the step of training the decision tree using a set of feature vectors and using the decision tree to mine non-linear association rules between features in the set of feature vectors to obtain a trained decision tree includes: S310, create an empty decision tree. The nodes of the decision tree represent features, the branches represent the values ​​of the features, the leaf nodes are used to represent the classification results, and the classification results represent the seismic phase type. S320, calculate the information gain ratio of each feature, select the feature with the largest information gain ratio as the splitting attribute of the current node, and generate multiple child nodes for the selected splitting attribute according to the discretization threshold. S330: Determine whether the generated child nodes meet the termination condition. If yes, execute S330; otherwise, execute S320. The termination condition is that the number of child node samples is less than the set quantity threshold or the purity reaches the set proportion threshold. S340 employs a pessimistic error pruning strategy to prune the constructed decision tree, forming classification rules from the root node to the leaf node, thus obtaining a trained decision tree.

5. The method according to claim 1, characterized in that, In S400, the step of training a Naive Bayes classifier using a set of feature vectors, calculating the global probability of seismic phase type using the Naive Bayes classifier, and obtaining a trained classifier includes: S410, perform data preprocessing on the feature vector set, dividing the feature vector set into a training set and a test set; S420: Based on the frequency of occurrence of features of various earthquake phases in the training set, calculate the conditional probability and prior probability of each feature under different earthquake phase categories. According to Bayes' theorem, multiply the prior probability and the conditional probability of each feature to obtain the posterior probability of each earthquake phase type. Based on the posterior probability of each earthquake phase type, obtain the global probability of the earthquake phase type. S430 uses the test set to evaluate the performance of the trained Naive Bayes classifier until the evaluation criteria are met, thus obtaining the trained classifier.

6. The method according to claim 1, characterized in that, In S500, the seismic phase identification based on the fusion model, and the output of the identification results of standard and anomalous seismic phases, include: S510 inputs newly acquired seismic wave data into the fusion model to obtain the class probability output by the decision tree and the posterior probability output by the Naive Bayes. S520 performs a weighted fusion of the class probability and the posterior probability to obtain the fused probability; the class with the highest probability is selected as the final classification result.

7. The method according to claim 1, characterized in that, Following S500, the method further includes: Set a confidence threshold for anomalous seismic phases. When the classification results of the seismic wave data by the trained decision tree and the trained classifier are consistent, and the weighted probability reaches the probability threshold, the seismic wave data is marked as a high-confidence anomalous signal.

8. A seismic phase identification system based on a classification algorithm, characterized in that, The system includes: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor performs the method as described in any one of claims 1 to 7.

Images