Lightweight earthquake monitoring method based on deep learning

Abstract

The invention discloses a lightweight earthquake monitoring method based on deep learning, and belongs to the technical field of earth science and deep learning, and the method comprises the steps: obtaining a data set for an earthquake monitoring task; constructing an earthquake monitoring model; wherein the earthquake monitoring model comprises a multi-scale feature extraction module, a feature encoder and a prediction layer; the multi-scale feature extraction module is used for extracting multi-dimensional features of seismic wave data; the feature encoder is used for modeling the multi-dimensional features extracted by the multi-scale feature extraction module; the prediction layer is used for outputting an earthquake monitoring task result based on the feature data output by the feature encoder; training the constructed earthquake monitoring model by using the data set; and realizing an earthquake monitoring task by using the trained earthquake monitoring model. According to the method, the problems of low precision and slow speed of the existing model method are effectively solved under the condition of realizing the light weight of the model.

Description

Technical Field

[0001] The present invention relates to the field of earth science and deep learning technology, and in particular to a lightweight earthquake monitoring method based on deep learning. Background Art

[0002] The detection of earthquake signals, phase picking and magnitude estimation are key tasks in earthquake monitoring. Earthquake detection is the identification of earthquake signals in the waveform recorded by seismic sensors, while phase picking is the process of determining the arrival time of different phases of earthquake waves by analyzing earthquake waveform signals, mainly including the phase picking of P waves (longitudinal waves) and S waves (transverse waves). This process is crucial for earthquake location and source analysis. Earthquake magnitude is a basic parameter that reflects the intensity of the source. It is used not only for scientific research, but also to provide the public with quick information on earthquakes to reduce casualties and property losses.

[0003] In recent years, researchers have been committed to the study of automated earthquake detection, phase picking, and magnitude estimation methods. The short-term average / long-term average (STA / LTA) method is often used in earthquake detection and phase picking. The signal is identified by comparing the short-term and long-term energy windows. The peak value above the threshold marks the arrival of the pulse P wave or S wave. Subsequent studies have proposed a variety of earthquake signal detection and identification methods by improving the STA / LTA method, combining autoregressive modeling and information criteria to determine the signal start time, and using neural networks and manual feature definitions to select P-wave and S-wave phases. The traditional magnitude estimation method is to estimate the magnitude based on the relationship between the maximum amplitude of the seismic wave and the distance from the epicenter. The specific method is to calculate based on the amplitude measured by the seismic recorder and combined with the earthquake distance correction table.

[0004] In recent years, deep learning technology has outperformed traditional methods in seismology and significantly improved performance. Therefore, deep learning has become more widely and deeply applied in seismological research, reflecting its high efficiency and stability. Seismic waveforms contain rich information that can be used to extract earthquake-related features and lay the foundation for subsequent analysis. With the increase of labeled seismic waveform data, deep learning models can capture deep features such as the phase and magnitude of P and S waves through training on these data. In the fields of earthquake detection, phase picking, and magnitude prediction, deep learning models have demonstrated powerful performance. The latest deep learning-based earthquake models, such as the multimodal earthquake model SeisCLIP for distinguishing explosions from earthquake events, the earthquake early warning model SeisParaNet, and the general model SeisT trained on large-scale datasets.

[0005] In recent years, methods based on convolutional neural networks have made many achievements in the field of earthquakes. PhaseNet uses three-component seismic waveforms as input and generates probability distributions of P-wave arrivals, S-wave arrivals, and noise as outputs, so that the peaks in the probability distribution provide accurate arrival times for P-waves and S-waves. However, although its picking accuracy is low, it still has deficiencies in computational complexity and inference speed. In 2019, Mousavi published Magnet, which is a deep learning model based on convolution applied to magnitude estimation tasks and can accurately estimate magnitude. Magnet models seismic waveforms through convolution and long short-term memory networks, but its results are still not accurate enough, and the training speed and inference speed are slow. In the earthquake detection task, CRED uses a combination of convolutional layers and bidirectional long short-term memory units in the residual structure, which can initially realize the detection of seismic signals, but its robustness and real-time performance in complex events need to be improved.

[0006] As a Transformer-based seismic wave model, EQTransformer has achieved good performance in earthquake detection and phase picking. Its architecture effectively captures the temporal dependencies in seismic data through the self-attention mechanism and can extract local and global features. Thanks to the Transformer framework, EQTransformer improves the picking accuracy and detection performance when dealing with complex scenarios. However, although the model performs well in complex scenarios, its computational complexity is high, resulting in a relatively slow running speed of the model, especially when dealing with large-scale data. SeisT is a comprehensive Transformer model designed specifically for tasks such as earthquake event detection, phase picking, and magnitude estimation. Although SeisT performs well in multi-task processing and global feature extraction, due to the high computational overhead of the self-attention mechanism, there is still room for improvement in the efficiency of the model in practical applications. In addition, although SeisT can capture long-term and short-term dependencies, its computational cost is still high when dealing with ultra-large-scale earthquake events, which limits its application in real-time monitoring.

[0007] In summary, most of the existing deep learning-based seismic wave methods are based on convolutional network structures and Transformer network structures. However, traditional convolutional networks cannot obtain the deep-level features of seismic waves, and thus cannot achieve high-precision prediction results; the computational complexity of the Transformer network structure increases significantly when the input sequence length increases, resulting in low efficiency when processing long-sequence seismic wave data. Summary of the invention

[0008] The present invention provides a lightweight earthquake monitoring method based on deep learning to solve the technical problems that the existing earthquake monitoring methods cannot obtain high-precision prediction results and have low efficiency.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0010] On the one hand, the present invention provides a lightweight earthquake monitoring method based on deep learning, comprising:

[0011] Obtain a data set for an earthquake monitoring task; wherein the sample data in the data set is seismic wave data;

[0012] Constructing an earthquake monitoring model; wherein the earthquake monitoring model comprises: a multi-scale feature extraction module, a feature encoder and a prediction layer; the multi-scale feature extraction module is used to extract multi-dimensional features of seismic wave data; the feature encoder is used to model the multi-dimensional features extracted by the multi-scale feature extraction module; the prediction layer is used to output earthquake monitoring task results based on the feature data output by the feature encoder;

[0013] Using the data set to train the constructed earthquake monitoring model;

[0014] Use the trained earthquake monitoring model to realize earthquake monitoring tasks.

[0015] Furthermore, the multi-scale feature extraction module combines basic convolution and spline convolution to use two different convolution layers to extract features from seismic wave data, thereby extracting multi-dimensional features of the seismic wave data.

[0016] Furthermore, the multi-scale feature extraction module includes: a linear normalization layer, a spline convolution layer, a base convolution layer, a fusion layer and a linear projection layer;

[0017] The input of the multi-scale feature extraction module is seismic wave data; the seismic wave data input to the multi-scale feature extraction module are respectively input to the linear normalization layer and the basic convolution layer; the output of the linear normalization layer is processed by a preset spline function and then input to the spline convolution layer, and the output of the spline convolution layer is input to the fusion layer; the output of the basic convolution layer is processed by a preset activation function and then input to the fusion layer; the output of the fusion layer is input to the linear projection layer, and the output of the linear projection layer is input to the feature encoder.

[0018] Furthermore, the process of processing the input seismic wave data by the multi-scale feature extraction module includes:

[0019] The input seismic wave data is linearly normalized through the linear normalization layer. The linear normalization result is processed by the preset spline function and then input into the spline convolution layer for spline convolution operation to obtain spline convolution features. The input seismic wave data is convolved through the basic convolution layer. The convolution result is processed by the preset activation function to obtain convolution features. The spline convolution features and convolution features are spliced ​​on the channel through the fusion layer to form a feature tensor. The feature tensor is linearly mapped through the linear projection layer, and then the linear mapping result is standardized using batch normalization.

[0020] Furthermore, the feature encoder includes: a linear projection layer, an adaptive convolution kernel layer, a Mamba layer, and a summation and normalization layer; wherein the linear projection layer, the adaptive convolution kernel layer, the Mamba layer, and the summation and normalization layer are connected in sequence; the adaptive convolution kernel layer includes a depth-separable convolution of multiple convolution kernels of different sizes.

[0021] Furthermore, the process of modeling the multidimensional features extracted by the multi-scale feature extraction module by the feature encoder includes:

[0022] A linear projection operation is performed on the input multidimensional features through a linear projection layer to project the input multidimensional features into multiple feature spaces; the feature data output by the linear projection layer is processed by using depthwise separable convolutions of convolution kernels of different sizes through the adaptive convolution kernel layer, and the processing results of the depthwise separable convolution are spliced ​​on channels and then batch normalized to obtain the output of the adaptive convolution kernel layer;

[0023] The output of the adaptive convolution kernel layer passes through the Mamba layer and the summing and normalization layer in turn to further encode the output of the adaptive convolution kernel layer and perform spatial modeling. The processing process is expressed as:

[0024] x ′ =σ(Conv1D(Linear(x)))

[0025] z=σ(Linear(x))

[0026]

[0027] y=LayerNorm(y ′ +x)

[0028] Among them, x is the input of the Mamba layer; Linear is a linear connection; Conv1D is a one-dimensional convolution; σ is an activation function; Represents element-by-element multiplication; SelectiveSSM represents the selective state space equation; LayerNorm represents layer normalization.

[0029] Furthermore, the earthquake monitoring task includes an earthquake event detection task, a seismic phase picking task and an earthquake magnitude estimation task.

[0030] Furthermore, in the earthquake event detection task and the seismic phase picking task, the prediction layer successively restores the feature map size through one-dimensional convolution, activation function and upsampling to match the scale of the input earthquake waveform, and finally outputs a probability distribution with the same sample scale; in the earthquake magnitude estimation task, the prediction layer successively simplifies the high-dimensional features into scalar outputs through global pooling, flattening and linear regression, and outputs the magnitude estimation results.

[0031] On the other hand, the present invention further provides an electronic device, comprising a processor and a memory; wherein the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the above method.

[0032] In yet another aspect, the present invention further provides a computer-readable storage medium, wherein at least one instruction is stored in the storage medium, and the instruction is loaded and executed by a processor to implement the above method.

[0033] The beneficial effects brought about by the technical solution provided by the present invention include at least:

[0034] The lightweight earthquake monitoring method based on deep learning of the present invention proposes an efficient lightweight earthquake monitoring model based on a state space model. The earthquake monitoring model network combines the advantages of the convolution module and the linear space model Mamb to perform multi-task modeling for earthquake waveform data. The fused model not only improves the ability to capture earthquake features, but also optimizes model accuracy, robustness training and reasoning speed. Thus, while achieving a lightweight model, the problems of low accuracy and slow speed of the existing model method are effectively solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 is a schematic diagram of the structure of a multi-scale spline convolution module provided in an embodiment of the present invention;

[0037] Figure 2 It is a schematic diagram of the overall structure of the SeisMamba model provided by an embodiment of the present invention;

[0038] Figure 3 is a schematic diagram of a prediction layer structure provided by an embodiment of the present invention;

[0039] Figure 4 It is a system block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0041] First of all, it should be noted that in the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "exemplarily" is intended to present the concept in a concrete way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0042] First embodiment

[0043] This embodiment provides a lightweight earthquake monitoring method based on deep learning, which can be implemented by an electronic device, which can be a terminal or a server. The execution process of the method includes the following steps:

[0044] S1, obtain the dataset for earthquake monitoring task;

[0045] Among them, the sample data in the data set is seismic wave data;

[0046] S2, construct an earthquake monitoring model, denoted as SeisMamba model;

[0047] The SeisMamba model includes: a multi-scale feature extraction module based on convolution, a feature encoder based on convolution and Mamba, and a prediction layer with multi-task output; the multi-scale feature extraction module is used to extract the multi-dimensional features of seismic wave data; the feature encoder is used to model the multi-dimensional features extracted by the multi-scale feature extraction module; the prediction layer is used to output the seismic monitoring task results based on the feature data output by the feature encoder;

[0048] S3, using the dataset to train the constructed earthquake monitoring model;

[0049] S4, use the trained earthquake monitoring model to realize the earthquake monitoring task.

[0050] Specifically, Figure 1As shown, the convolution-based multi-scale feature extraction module in this embodiment is a multi-scale spline convolution module, which includes: a linear normalization layer, a spline convolution layer, a basic convolution layer, a fusion layer and a linear projection layer; the input of the multi-scale feature extraction module is seismic wave data; the seismic wave data input to the multi-scale feature extraction module are respectively input to the linear normalization layer and the basic convolution layer; the output of the linear normalization layer is processed by a preset radial basis function and then input to the spline convolution layer, and the output of the spline convolution layer is input to the fusion layer; the output of the basic convolution layer is processed by a preset activation function SiLu and then input to the fusion layer; the output of the fusion layer is input to the linear projection layer, and the output of the linear projection layer is input to the feature encoder.

[0051] The multi-scale feature extraction module combines basic convolution and spline convolution, and uses two different convolution layers to extract features from seismic waves, which improves the feature extraction capability of the input signal. Its internal structure is as follows: Figure 1 The core of the proposed method is to use a spline convolution layer, use radial basis functions to perform nonlinear mapping on the input data, and combine convolution operations to achieve efficient encoding of seismic waveforms. Specifically, the multi-scale feature extraction module enhances the performance of feature extraction by mapping the input signal to different feature spaces, especially when processing complex seismic waveforms. Set the input of the multi-scale feature extraction module to C i Indicates the number of channels of the i-th layer input, L i represents the sequence length; t represents the total number of convolution modules, and B represents the number of batches. The operation of the multi-scale feature extraction module can be formally expressed as:

[0052]

[0053] in, Represents the output y of different convolution modules i Perform a concatenation operation on the channel. This step combines the outputs of all modules to form a tensor containing rich features. Then proj represents a linear mapping of the results. Next, batch normalization is used to standardize the concatenated output to improve the stability of training and the generalization ability of the model. Figure 1 As shown in the figure, the multi-scale feature extraction module maps the seismic wave signal to a low-dimensional feature space through the convolution operation in the time dimension, while increasing the depth channel of the feature, so that the key information can be retained and the characteristic information of the seismic wave can be extracted more efficiently. This module contains ordinary convolution layers and spline convolution layers, and the outputs of different convolution layers are spliced ​​on the channel. This step combines the outputs of all modules to form a tensor containing rich features. The results are then linearly mapped and standardized to improve the stability of training and the generalization ability of the model.

[0054] Figure 2 The overall architecture of SeisMamba of the present invention is shown. The encoder designed in this embodiment is an encoder based on state space model and convolution fusion, and its structure is divided into two sequentially executed modules. The specific structure is as follows Figure 2 The right side of the network includes a linear projection layer, an adaptive convolution kernel layer, a Mamba layer, and an addition and normalization layer connected in sequence. First, an efficient and adaptive convolution module is designed, called an adaptive convolution kernel layer, and it is used as the encoder part. It uses adaptive convolution kernels to fuse features through convolution kernels of different sizes. This part projects the shallow features of the input into multiple feature spaces through linear projection. This method enables the model to consider feature space information from different scales and positions, and promotes effective learning of representations. The proposed adaptive convolution kernel layer is defined as follows:

[0055]

[0056] Among them, s i is the output of the depthwise separable convolution, and its expression formula is as follows:

[0057] S i =DSC(BN(W i *X+b i ))+proj(X) (3)

[0058] Among them, BN stands for batch normalization; W i represents the weight matrix of the linear projection layer; X represents the input of the adaptive convolution kernel layer; b i represents the bias matrix of the linear projection layer; DSC represents depthwise separable convolution;

[0059] The core part of the Mamba encoder is the Mamba layer, which further encodes and spatially models the features obtained from the previous module. This layer is designed based on the idea of ​​the state-space model. Mamba further introduces a selective mechanism to filter out irrelevant information and designs a hardware-aware algorithm to improve its computational efficiency. This improvement is particularly suitable for modeling seismic data, demonstrating strong modeling capabilities and scalability. Through this selective mechanism, the Mamba layer can extract effective information from the input data in a targeted manner, significantly improving the performance of the model on complex sequence data. Let x be the input of the mamba layer, the process can be described as:

[0060]

[0061] Where σ is the activation function SiLU. Then, x, x ′ And z is used to calculate the output y, Conv1D is a one-dimensional convolution, and Linear is a linear connection:

[0062]

[0063] y=LayerNorm(y ′ +x) (6)

[0064] Among them, SelectiveSSM represents the selective state space equation; LayerNorm represents layer normalization; Represents element-by-element multiplication. Mamba Layer demonstrates excellent performance in seismic data analysis by combining the long-term dependency modeling capability of the state-space model with the selective information filtering mechanism. It also has good hardware adaptability and is suitable for efficient processing of large-scale seismic data.

[0065] Furthermore, this embodiment uses multiple prediction layers to cope with different tasks. Figure 3 The prediction layer in generates detection or estimation results according to different tasks. In the earthquake event detection and phase picking tasks, the prediction layer restores the feature map size through one-dimensional convolution, activation function and upsampling to match the scale of the input earthquake waveform, and finally outputs a probability distribution with the same sample scale. In regression tasks such as earthquake magnitude estimation, the prediction layer simplifies high-dimensional features into scalar outputs through global pooling and linear regression to generate magnitude estimation results.

[0066] The data processing process of the above model is as follows: first, the three-part seismic wave data is input into the model to obtain the multi-dimensional features of the input waveform, then the features are modeled through the Mamba encoder, and finally different results are output according to different tasks through the prediction layer, and the regression loss and cross entropy loss are calculated with the label, and finally R is calculated. 2 , precision (Precision), recall (Re), F1 score and other evaluation indicators.

[0067] The above model is trained and tested on the Diting dataset, and the results are shown in Table 1.

[0068] Table 1 Test results of the model on each task on the Diting dataset

[0069]

[0070] As can be seen from the above table, the proposed method performs earthquake detection, phase picking and magnitude estimation tasks on the Diting dataset; the earthquake detection task achieves an accuracy of 97.29 and a recall rate of 97.85; the P wave in the phase picking task achieves an accuracy of 94.07 and an F1 score of 92.20; the S phase picking achieves an accuracy of 90.52 and an F1 score of 84.16; in the magnitude estimation task, this method achieves an R of 0.915. 2 .

[0071] In summary, in order to solve the problem that the current deep learning models in the earthquake field lack lightweight, this embodiment proposes an efficient and lightweight earthquake monitoring model based on the state space model. This earthquake monitoring model has demonstrated powerful performance in tasks such as earthquake detection, seismic phase picking, and magnitude estimation, and has great application prospects.

[0072] Second embodiment

[0073] This embodiment provides an electronic device, such as Figure 4 As shown, the electronic device includes: a processor and a memory; wherein the processor and the memory can be connected via a communication bus; the memory stores at least one instruction, and the instruction is loaded and executed by the processor to implement the method of the first embodiment. In addition, the electronic device may also include a transceiver, the processor and the transceiver can be connected via a communication bus, and the transceiver is used to communicate with other devices.

[0074] Next, combine Figure 4 The following is a detailed introduction to the various components of the electronic device:

[0075] Among them, the processor is the control center of the electronic device, and the electronic device may include multiple processors, each of which may be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The processor here may be a processor or a general term for multiple processing elements. For example, the processor is one or more central processing units (CPUs), or other general-purpose processors, application specific integrated circuits (ASICs), or one or more integrated circuits configured to implement an embodiment of the present invention, such as one or more microprocessors (digital signal processors, DSPs), or one or more field programmable gate arrays (field programmable gate arrays, FPGAs), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor may execute various functions of the electronic device by running or executing software programs stored in the memory and calling data stored in the memory.

[0076] In a specific implementation, as an embodiment, the processor may include one or more CPUs, such as Figure 4The CPU0 and CPU1 shown in the figure are, of course, only exemplary.

[0077] The memory is used to store the software program for executing the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can refer to the above method embodiment and will not be repeated here.

[0078] Optionally, the memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory may be integrated with the processor or exist independently and accessed through the interface circuit ( Figure 4 (not shown) is coupled to the processor, which is not specifically limited in this embodiment of the present invention.

[0079] The transceiver may include a receiver and a transmitter ( Figure 4 The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function. The transceiver can be integrated with the processor or exist independently and communicate with the electronic device through the interface circuit ( Figure 4 (not shown) is coupled to the processor, which is not specifically limited in this embodiment of the present invention.

[0080] In addition, it should be noted that Figure 4 The structure of the electronic device shown in the figure does not constitute a limitation on the device, and the actual device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. In addition, the technical effects achieved by the electronic device when executing the method of the first embodiment above can refer to the technical effects described in the first embodiment above, so they are not repeated here.

[0081] Third embodiment

[0082] This embodiment provides a computer-readable storage medium, which stores at least one instruction, and the instruction is loaded and executed by a processor to implement the method of the first embodiment. The computer-readable storage medium may be a ROM, a random access memory, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc. The instructions stored therein may be loaded by a processor in a terminal to execute the method.

[0083] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiment of the present invention can be in the form of a full or partial hardware embodiment, a full or partial software embodiment or an embodiment combining software and hardware. Moreover, when implemented using software, the embodiment of the present invention can be in the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program codes. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center containing one or more available media sets. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium. The semiconductor medium may be a solid state hard disk.

[0084] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0085] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0086] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal 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, article or terminal device. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements. In addition, the term "and / or" is only an association relationship describing the associated objects, indicating that there can be three relationships, for example, A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone, wherein A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding. "At least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can be represented by: a, b, c, ab, ac, bc or abc, where a, b, c can be single or plural.

[0087] In addition, it can be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0088] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0089] In several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of functional modules / units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point, the coupling or direct coupling or communication connection between each other shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms. The unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, each functional unit in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0090] If the method is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0091] Finally, it should be noted that the above is only a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for ordinary technicians in this technical field, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A lightweight earthquake monitoring method based on deep learning, characterized in that: include: Obtain a data set for an earthquake monitoring task; wherein the sample data in the data set is seismic wave data; Constructing an earthquake monitoring model; wherein the earthquake monitoring model comprises: a multi-scale feature extraction module, a feature encoder and a prediction layer; the multi-scale feature extraction module is used to extract multi-dimensional features of seismic wave data; the feature encoder is used to model the multi-dimensional features extracted by the multi-scale feature extraction module; the prediction layer is used to output earthquake monitoring task results based on the feature data output by the feature encoder; Using the data set to train the constructed earthquake monitoring model; Use the trained earthquake monitoring model to realize earthquake monitoring tasks.

2. The lightweight earthquake monitoring method based on deep learning according to claim 1, characterized in that: The multi-scale feature extraction module combines basic convolution and spline convolution, uses two different convolution layers to extract features from seismic wave data, and extracts multi-dimensional features of the seismic wave data.

3. The lightweight earthquake monitoring method based on deep learning according to claim 2, characterized in that: The multi-scale feature extraction module includes: a linear normalization layer, a spline convolution layer, a base convolution layer, a fusion layer and a linear projection layer; The input of the multi-scale feature extraction module is seismic wave data; the seismic wave data input to the multi-scale feature extraction module are respectively input to the linear normalization layer and the basic convolution layer; the output of the linear normalization layer is processed by a preset spline function and then input to the spline convolution layer, and the output of the spline convolution layer is input to the fusion layer; the output of the basic convolution layer is processed by a preset activation function and then input to the fusion layer; the output of the fusion layer is input to the linear projection layer, and the output of the linear projection layer is input to the feature encoder.

4. The lightweight earthquake monitoring method based on deep learning according to claim 3, characterized in that: The process of the multi-scale feature extraction module processing the input seismic wave data includes: The input seismic wave data is linearly normalized through the linear normalization layer. The linear normalization result is processed by the preset spline function and then input into the spline convolution layer for spline convolution operation to obtain spline convolution features. The input seismic wave data is convolved through the basic convolution layer. The convolution result is processed by the preset activation function to obtain convolution features. The spline convolution features and convolution features are spliced ​​on the channel through the fusion layer to form a feature tensor. The feature tensor is linearly mapped through the linear projection layer, and then the linear mapping result is standardized using batch normalization.

5. The lightweight earthquake monitoring method based on deep learning according to claim 1, characterized in that: The feature encoder includes: a linear projection layer, an adaptive convolution kernel layer, a Mamba layer, and a summation and normalization layer; wherein the linear projection layer, the adaptive convolution kernel layer, the Mamba layer, and the summation and normalization layer are connected in sequence; the adaptive convolution kernel layer includes a depth-separable convolution of multiple convolution kernels of different sizes.

6. The lightweight earthquake monitoring method based on deep learning according to claim 5, characterized in that: The process of modeling the multidimensional features extracted by the multi-scale feature extraction module by the feature encoder includes: A linear projection operation is performed on the input multidimensional features through a linear projection layer to project the input multidimensional features into multiple feature spaces; the feature data output by the linear projection layer is processed by using depthwise separable convolutions of convolution kernels of different sizes through the adaptive convolution kernel layer, and the processing results of the depthwise separable convolution are spliced ​​on channels and then batch normalized to obtain the output of the adaptive convolution kernel layer; The output of the adaptive convolution kernel layer passes through the Mamba layer and the summing and normalization layer in turn to further encode the output of the adaptive convolution kernel layer and perform spatial modeling. The processing process is expressed as: x ′ =σ(Conv1D(Linear(x))) z=σ(Linear(x)) y=LayerNorm(y ′ +x) Among them, x is the input of the Mamba layer; Linear is a linear connection; Conv1D is a one-dimensional convolution; σ is an activation function; Represents element-by-element multiplication; SelectiveSSM represents the selective state space equation; LayerNorm represents layer normalization.

7. The lightweight earthquake monitoring method based on deep learning according to claim 1, characterized in that: The earthquake monitoring task includes an earthquake event detection task, a seismic phase picking task and an earthquake magnitude estimation task.

8. The lightweight earthquake monitoring method based on deep learning according to claim 7, characterized in that: In the earthquake event detection task and the seismic phase picking task, the prediction layer successively restores the feature map size through one-dimensional convolution, activation function and upsampling to match the scale of the input earthquake waveform, and finally outputs a probability distribution with the same sample scale; in the earthquake magnitude estimation task, the prediction layer successively simplifies high-dimensional features into scalar outputs through global pooling, flattening and linear regression, and outputs the magnitude estimation result.

Images