CN118244340B - Earthquake event type identification judging method based on deep learning - Google Patents

Abstract

The invention relates to the technical field of earthquake identification, and discloses a method for identifying and judging earthquake event types based on deep learning; collecting real-time event characteristic data, inputting the real-time event characteristic data into a machine learning model after training to predict a real-time event type value, collecting negative influence parameters of a natural seismic event, generating natural land earthquake shadow loudness, comparing natural land earthquake shadow loudness with a preset natural seismic influence threshold value, dividing seismic influence levels, and generating a seismic influence early warning prompt; compared with the prior art, the method has the advantages that the event characteristic data can be accurately acquired, the type of the earthquake event can be accurately predicted, when the natural earthquake event is identified, the damage severity of the natural earthquake event is adaptively calculated and estimated, and the damage early warning prompt of the natural earthquake event is given by combining the estimation result, so that a good foundation is provided for the manager of the earthquake station to quickly, accurately and timely send out the earthquake early warning information.

Description

Earthquake event type identification judging method based on deep learning

Technical Field

The invention relates to the technical field of seismic identification, in particular to a seismic event type identification judging method based on deep learning.

Background

The earthquake monitoring is to monitor and measure the earthquake precursor abnormality and the earthquake activity before and after the earthquake, and because the mining field needs to operate in a mode of detonating explosive underground, when the explosion power is high, the phenomenon of the earthquake event in a local area can be caused, a certain interference is caused to the identification of the earthquake type, and in order to realize the effective monitoring of the natural earthquake, the identification of whether the earthquake occurs is needed, so that the specific type of the earthquake event is judged.

The Chinese patent with the application publication number of CN108387932B discloses a method and a device for identifying seismic signals, which are characterized in that a total seismic channel is generated according to at least one seismic channel, a main seismic signal and at least one signal to be identified are determined from the total seismic channel, the similarity between each signal to be identified and the main seismic signal is determined, the seismic signals in the signals to be identified are determined according to the similarity, when the similarity is higher, the seismic signals can be identified, and the seismic signals can be identified more accurately through the similarity determination in the process;

The prior art has the following defects:

When the existing earthquake event type is identified and judged, the earthquake event type is identified by comparing the similarity between the signal to be identified and the main earthquake signal, and after the earthquake event type is identified, no targeted calculation analysis is performed on the hazard severity of the natural earthquake event, so that the hazard severity of the natural earthquake event cannot be judged quickly and accurately, the adaptive early warning prompt cannot be performed according to the actual earthquake condition, and the timely generation and release of the natural earthquake event early warning information are not facilitated.

In view of the above, the present invention proposes a method for determining seismic event type identification based on deep learning to solve the above-mentioned problems.

Disclosure of Invention

In order to overcome the defects in the prior art, the invention provides the following technical scheme for achieving the purposes: the seismic event type identification judging method based on deep learning comprises the following steps:

s1: collecting historical training data of the seismic event, wherein the historical training data comprises event characteristic data and event type values;

S2: training a machine learning model for predicting an event type value based on historical training data;

S3: collecting real-time event characteristic data, inputting the real-time event characteristic data into a machine learning model after training to predict real-time event type values, wherein the event type values comprise natural seismic events and non-natural seismic events;

S4: collecting negative influence parameters of a natural seismic event, and generating natural ground earthquake shadow loudness based on the negative influence parameters;

S5: and comparing the loudness of the natural land earthquake shadow with a preset natural earthquake influence threshold value, dividing earthquake influence levels, and generating earthquake influence early warning prompts.

Further, the event characteristic data comprise an aspect ratio difference value, a frequency vibration value, a residual vibration density and a primary compression difference value;

The method for acquiring the aspect ratio difference value comprises the following steps:

screening out with the center of the seismic source as the starting point and the length greater than a preset safety distance as the standard The earthquake stations are marked as target earthquake stations;

Acquisition of The occurrence times of longitudinal waves and transverse waves monitored by each target seismic station in a preset acquisition period are respectively obtainedLongitudinal wave valuesA transverse wave value;

Will be The longitudinal wave values are respectively compared with the time length corresponding to the preset acquisition period to obtainSub-longitudinal wave values;

The expression of the sub-longitudinal wave value is:

；

In the method, in the process of the invention, Is the firstThe number of sub-longitudinal wave values,Is the firstThe number of longitudinal wave-length values,The method comprises the steps of presetting a duration corresponding to a collection period;

Will be The transverse wave values are respectively compared with the time length corresponding to the preset acquisition period to obtainSub-transverse wave values;

The expression of the sub-transverse wave value is:

；

In the method, in the process of the invention, Is the firstThe value of the sub-transverse wave,Is the firstA transverse wave value;

Removing maximum value and minimum value of sub longitudinal wave value and sub transverse wave value respectively, and keeping the rest After the summation of the sub-longitudinal wave values, the sub-longitudinal wave values are combined withComparing the accumulated values of the sub-transverse wave values to obtain a longitudinal-transverse wave ratio;

The expression of the longitudinal-transverse wave ratio is:

；

In the method, in the process of the invention, In order to achieve an aspect ratio of the waves,Is the firstThe number of sub-longitudinal wave values,Is the firstSub-transverse wave values;

After the longitudinal-transverse wave ratio is subjected to difference with the standard ratio, an aspect ratio difference value is obtained;

the expression of the aspect ratio difference is:

；

In the method, in the process of the invention, Is the difference in the aspect ratio,Is a standard ratio.

Further, the method for acquiring the frequency vibration value comprises the following steps:

Measuring the distance between a target seismic station and the center of a seismic source, marking the target seismic station corresponding to the minimum distance value as a seismic station to be detected, and monitoring the seismic wave oscillation frequency of the seismic station to be detected through a seismograph;

The time when the vibration frequency of the earthquake wave reaches the preset lower limit frequency for the first time is recorded as the starting time, and the time when the vibration frequency of the earthquake wave reaches the preset upper limit frequency for the first time is recorded as the ending time;

after the end time and the start time are differenced, an oscillation period is obtained;

The expression of the oscillation period is:

；

In the method, in the process of the invention, For the duration of the oscillation,In order to finish the moment of time,Is the starting time;

Randomly marking in the oscillation period by taking the time length corresponding to one tenth of the oscillation period as a standard Each of the sub-sections not adjacent to each other are countedTotal number of oscillations within the subsection;

Will be The total times of oscillation in the subsections are respectively compared with the time length corresponding to the subsections to obtainSub-frequencies;

the expression of the sub-frequencies is:

；

In the method, in the process of the invention, Is the firstThe sub-frequencies of the frequency band are selected,Is the firstTotal number of oscillations within the subsection;

Will be The sub-frequencies are accumulated and averaged to obtain a frequency vibration value;

the expression of the frequency vibration value is:

；

In the method, in the process of the invention, As the value of the frequency oscillation,Is the firstSub-frequencies.

Further, the method for acquiring the residual shock intensity comprises the following steps:

taking the moment when the seismograph monitors the earthquake waves for the first time as a first moment and the moment when the seismograph monitors the earthquake waves for the last time as a second moment, and measuring the vibration amplitudes of all the earthquake waves from the first moment value to the second moment;

Marking the time corresponding to the maximum value of the vibration amplitude of the seismic wave as a target time, and marking the target time to a second time as an effective period;

Equally dividing the effective period into Sub-time periods, respectively countThe number of seismic waves monitored during a sub-period is obtainedThe residual shock value;

Will be The aftershock values are compared with the time length corresponding to the subinterval to obtainSub-densities;

The expression of the sub-density is:

；

In the method, in the process of the invention, Is the firstThe degree of sub-concentration is set to,Is the firstThe residual shock value is a value of the residual shock,The time length corresponds to the sub-period;

Based on Generating the sub-intensity and the aftershock intensity;

The expression of the residual shock concentration is:

；

In the method, in the process of the invention, For the concentration degree of the aftershocks,Is the firstSub-densities.

Further, the method for obtaining the initial compression difference value comprises the following steps:

Drawing a circle by taking the center of the seismic source as the center of a circle and taking the lengths of three preset safety distances as the radius to obtain a monitoring circle, drawing a line in the monitoring circle after passing through the center of the circle, and equally dividing the monitoring circle into A plurality of regions;

monitoring by seismometer The initial vibration direction of the seismic waves in all the target seismic stations in the individual areas;

The seismic wave with the initial vibration direction of outward compression is recorded as compression wave, the seismic wave with the initial vibration direction of inward expansion is recorded as expansion wave, and statistics is carried out Total sum of compression waves of individual regionsThe total amount of expansion waves for each region;

Will be Total sum of compression waves of individual regionsThe total amount of the expansion waves of each region is subjected to difference comparison one by one to obtainSub-compression differences;

the expression of the sub-compression difference is:

；

In the method, in the process of the invention, Is the firstThe sub-compressed difference value is used to determine,Is the firstThe total amount of compressional waves in the individual zones,Is the firstThe total amount of expansion waves for each region;

the minimum value of the sub-compression difference value is subjected to difference with a preset standard compression difference value, and an initial compression difference value is obtained;

the expression of the initial compression difference is:

；

In the method, in the process of the invention, For the initial compression difference value,Is the minimum value of the sub-compressed difference value,Is the standard compression difference.

Further, the training method of the machine learning model for predicting the event type value comprises the following steps:

Converting the event feature data into a corresponding set of feature vectors;

And taking the feature vector as input of a machine learning model, taking event type values corresponding to each group of event feature data as output of the machine learning model, taking the event type values as prediction targets, taking the sum of prediction errors of all the minimized training data as a training target, and training the machine learning model until the sum of the prediction errors reaches convergence, and stopping training.

Further, when the event type value is a natural seismic event, the output of the machine learning model is 1;

When the event type value is a non-natural seismic event, the output of the machine learning model is 0.

Further, the negative impact parameters include seismic magnitude and duration ratio:

the method for acquiring the earthquake magnitude value comprises the following steps:

drawing a circle by taking the center of the seismic source as a starting point and taking the length 100 kilometers away from the starting point as a radius to obtain a screening circle;

Marking the earthquake station on the screening circle, and connecting a starting point and a terminal point by taking the position of the center of the urban area as the terminal point to obtain a magnitude line;

when the earthquake station is arranged at the superposition position of the magnitude line and the screening circle, marking the earthquake station at the superposition position of the magnitude line and the screening circle as an available earthquake station;

When no seismic station exists at the overlapping position of the magnitude line and the screening circle, marking the seismic station which is on the screening circle and has the smallest distance from the overlapping position of the magnitude line and the screening circle as the available seismic station;

Obtaining the maximum amplitude of the seismic wave received at 100 km from the center of the seismic source and the maximum amplitude of the seismic wave of the 0-level seismic wave received at 100 km from the center of the seismic source through the available seismic station, and generating a seismic magnitude value;

the expression of the seismic magnitude value is:

；

In the method, in the process of the invention, For the magnitude of the earthquake,For the maximum amplitude of the received seismic waves 100 km from the center of the source,Is the maximum amplitude of the seismic wave of a class 0 seismic received 100 km from the center of the source.

Further, the method for acquiring the duration ratio comprises the following steps:

random selection within a screening circle Individual seismic stations and acquiringThe value of the acceleration sensor in each seismic station;

Recording Obtaining the time period from the occurrence of the fluctuation to the no-fluctuation of the value of the acceleration sensor in each seismic stationTotal time length of each acceleration change;

Will be After the acceleration periods are accumulated, the total time length of the acceleration changes is compared to obtainA sub-occupation ratio;

the expression of the sub-ratio is:

；

In the method, in the process of the invention, Is the firstThe ratio of the sub-occupation ratios,Is the firstThe first seismic stationThe time period of the acceleration is a period,Is the firstThe total time length of the acceleration change of each seismic station;

Will be Averaging after accumulating the sub-occupation ratios to obtain a duration occupation ratio;

the duration ratio is expressed as:

；

In the method, in the process of the invention, In order for the duration to be a fraction of the time,Is the firstA sub-occupation ratio;

The expression for natural earthquake shadow loudness is:

；

In the method, in the process of the invention, In order to naturally be earthquake shadow loudness degrees,Is a natural constant which is used for the preparation of the composite material,、Is a weight factor.

Further, the earthquake influence level comprises a first-level early warning level, a second-level early warning level and a third-level early warning level;

the method for dividing the primary early warning level, the secondary early warning level and the tertiary early warning level comprises the following steps:

Natural seismic influence degree With a preset first natural seismic impact thresholdAnd a second natural seismic impact thresholdIn comparison with the comparison result of the comparison,Less than；

When (when)Less than or equal toDividing into first-level early warning levels;

When (when) Less thanAnd (2) andLess than or equal toDividing into two-level early warning levels;

When (when) Greater thanDividing into three levels of early warning;

The method for generating the earthquake influence early warning prompt comprises the following steps:

when the earthquake influence level is a first-level early warning level, generating no earthquake influence early warning prompt;

When the earthquake influence level is a secondary early warning level, generating an earthquake influence early warning prompt;

And when the earthquake influence level is three-level early warning level, generating an earthquake influence early warning prompt.

The method for identifying and judging the seismic event type based on deep learning has the technical effects and advantages that:

According to the method, historical training data of a seismic event are collected, the historical training data comprise event feature data and event type values, a machine learning model for predicting the event type values is trained based on the historical training data, real-time event feature data are collected, real-time event type values are predicted in the trained machine learning model, negative influence parameters of the natural seismic event are collected, natural land earthquake shadow loudness is generated based on the negative influence parameters, natural land earthquake shadow loudness is compared with a preset natural seismic influence threshold value, seismic influence levels are divided, and seismic influence early warning prompts are generated; compared with the prior art, the method has the advantages that the event characteristic data can be accurately acquired, the application range of the prediction recognition data of the earthquake event types is enlarged, the machine learning model can accurately predict the earthquake event types on the basis of diversified data, when the natural earthquake event is recognized, the damage severity of the natural earthquake event is adaptively calculated and evaluated, and the damage early warning prompt of the natural earthquake event is given by combining the evaluation result, so that a good basis is provided for the manager of the earthquake station to quickly, accurately and timely send out the earthquake early warning information.

Drawings

FIG. 1 is a schematic flow chart of a method for determining the type of seismic event based on deep learning according to embodiment 1 of the present invention;

fig. 2 is a schematic diagram of a seismic event type identification determination system based on deep learning according to embodiment 2 of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Example 1: referring to fig. 1, the method for determining the type of the seismic event based on deep learning according to the embodiment includes:

s1: collecting historical training data of the seismic event, wherein the historical training data comprises event characteristic data and event type values;

The event characteristic data is comprehensive data of relevant geology and waveforms generated when the earthquake event occurs and recorded by the earthquake station, the data change condition brought by the earthquake event can be recorded by collecting the event characteristic data and the event type value, and the event type value matched with the event characteristic data is obtained based on the recorded event characteristic data;

The event characteristic data comprise an aspect ratio difference value, a frequency vibration value, a residual vibration concentration and an initial compression difference value; the difference of the aspect ratio refers to the difference between the ratio of the longitudinal waves and the transverse waves generated when the earthquake event occurs and the standard ratio, the ratio of the number of the longitudinal waves to the number of the transverse waves in unit time can be reflected by collecting the ratio of the longitudinal waves and the transverse waves, and the number of the transverse waves generated by the earthquake is more than the number of the longitudinal waves because the longitudinal waves can pass through solid and liquid at a higher speed and the transverse waves can pass through liquid at a lower speed and are difficult to pass through liquid, and the larger the difference of the aspect ratio is, the larger the number of the longitudinal waves appears in the earthquake event is;

The method for acquiring the aspect ratio difference value comprises the following steps:

screening out with the center of the seismic source as the starting point and the length greater than a preset safety distance as the standard The earthquake stations are marked as target earthquake stations; the length of the preset safety distance is the minimum distance between two adjacent seismic stations for monitoring longitudinal waves and transverse waves in a seismic event, so that a certain distance is kept between the two adjacent seismic stations, the phenomenon that the number of the longitudinal waves and the transverse waves monitored in the two seismic stations with the too small distance is similar or even the same is avoided, and the independence of data in each seismic station is ensured; the length of the preset safety distance is obtained through coefficient optimization after acquiring a great number of minimum distance values of two seismic stations which can acquire complete and independent longitudinal wave and transverse wave data;

Acquisition of The occurrence times of longitudinal waves and transverse waves monitored by each target seismic station in a preset acquisition period are respectively obtainedLongitudinal wave valuesA transverse wave value; the preset acquisition period is a time length capable of meeting the quantity and sufficient acquisition of longitudinal waves and transverse waves, so that after one preset acquisition period, sufficient longitudinal wave and transverse wave data can be acquired in two adjacent seismic stations, and the data quantity is ensured to be sufficient; the method comprises the steps of presetting a collection period, wherein the collection period is obtained through repeated debugging after collecting a large number of enough longitudinal waves and transverse waves of a historical collection time corresponding to the number of the longitudinal waves and the transverse waves and combining actual collection operation parameters of a seismic instrument in a seismic station;

Will be The longitudinal wave values are respectively compared with the time length corresponding to the preset acquisition period to obtainSub-longitudinal wave values;

The expression of the sub-longitudinal wave value is:

；

In the method, in the process of the invention, Is the firstThe number of sub-longitudinal wave values,Is the firstThe number of longitudinal wave-length values,The method comprises the steps of presetting a duration corresponding to a collection period;

Will be The transverse wave values are respectively compared with the time length corresponding to the preset acquisition period to obtainSub-transverse wave values;

The expression of the sub-transverse wave value is:

；

In the method, in the process of the invention, Is the firstThe value of the sub-transverse wave,Is the firstA transverse wave value;

Removing maximum value and minimum value of sub longitudinal wave value and sub transverse wave value respectively, and keeping the rest After the summation of the sub-longitudinal wave values, the sub-longitudinal wave values are combined withComparing the accumulated values of the sub-transverse wave values to obtain a longitudinal-transverse wave ratio; the mode of deleting the maximum value and the minimum value can avoid deviation caused by irregular data acquisition or inaccurate data calculation during acquisition or calculation, so that the accuracy of subsequent data calculation is improved;

The expression of the longitudinal-transverse wave ratio is:

；

In the method, in the process of the invention, In order to achieve an aspect ratio of the waves,Is the firstThe number of sub-longitudinal wave values,Is the firstSub-transverse wave values;

after the longitudinal-transverse wave ratio is subjected to difference with the standard ratio, an aspect ratio difference value is obtained; the standard ratio is a normal set value of the longitudinal-transverse wave ratio when the longitudinal-transverse wave ratio is actually calculated, and is used for representing the numerical value of the longitudinal-transverse wave ratio and effectively partitioning the actual value of the longitudinal-transverse wave ratio; the standard ratio is obtained by averaging the aspect ratio obtained on the basis of no deviation of a large number of seismic instruments with the acquisition history and accurate data acquisition;

the expression of the aspect ratio difference is:

；

In the method, in the process of the invention, Is the difference in the aspect ratio,Is a standard ratio;

the frequency vibration value refers to the vibration frequency of the earthquake wave generated in unit time when the earthquake event occurs, and in unit time, the high-frequency wave has more vibration than the low-frequency wave, but the non-natural earthquake can generate more high-frequency waves than the natural earthquake, and when the frequency vibration value is larger, the vibration frequency of the earthquake wave in unit time is larger, the high-frequency waves are more;

The method for acquiring the frequency vibration value comprises the following steps:

Measuring the distance between a target seismic station and the center of a seismic source, marking the target seismic station corresponding to the minimum distance value as a seismic station to be detected, and monitoring the seismic wave oscillation frequency of the seismic station to be detected through a seismograph; by selecting the target seismic station closest to the center of the seismic source, the monitoring point closest to the center of the seismic source can be obtained, so that when the seismic waves appear and are conducted outwards, the transmission distance before the seismic waves are monitored is reduced, the attenuation degree of the vibration times of the seismic waves in the conduction is reduced, and the accuracy of the vibration frequency of the seismic waves is improved;

The time when the vibration frequency of the earthquake wave reaches the preset lower limit frequency for the first time is recorded as the starting time, and the time when the vibration frequency of the earthquake wave reaches the preset upper limit frequency for the first time is recorded as the ending time; the preset lower limit frequency and the preset upper limit frequency are used for limiting the minimum value and the maximum value of the vibration frequency of the earthquake waves, so that the vibration frequency of the earthquake waves is limited in a certain range, and the problems that the vibration frequency of the earthquake waves is not easy to monitor due to too small vibration frequency of the earthquake waves and the monitoring of the vibration frequency of the earthquake waves is inaccurate are avoided; the preset lower limit frequency and the preset upper limit frequency are obtained by acquiring a great number of historical minimum values and maximum values of the vibration frequencies of the seismic waves which are accurately monitored, and then obtaining the average value of the minimum values and the maximum values;

after the end time and the start time are differenced, an oscillation period is obtained;

The expression of the oscillation period is:

；

In the method, in the process of the invention, For the duration of the oscillation,In order to finish the moment of time,Is the starting time;

Randomly marking in the oscillation period by taking the time length corresponding to one tenth of the oscillation period as a standard Each of the sub-sections not adjacent to each other are countedTotal number of oscillations within the subsection; the random marking mode can achieve the effect of random selection, improves the randomness of data acquisition, simultaneously avoids mutual adhesion of oscillation data between adjacent subsections, and improves the autonomy of the oscillation data;

Will be The total times of oscillation in the subsections are respectively compared with the time length corresponding to the subsections to obtainSub-frequencies;

the expression of the sub-frequencies is:

；

In the method, in the process of the invention, Is the firstThe sub-frequencies of the frequency band are selected,Is the firstTotal number of oscillations within the subsection;

Will be The sub-frequencies are accumulated and averaged to obtain a frequency vibration value;

the expression of the frequency vibration value is:

；

In the method, in the process of the invention, As the value of the frequency oscillation,Is the firstSub-frequencies;

The aftershock concentration refers to the degree of tightness of a smaller earthquake wave after a larger earthquake wave occurs in the time of occurrence of an earthquake event, when a natural earthquake occurs, the natural earthquake can change the rock structure around a stress field to cause new cracks or slide along the existing cracks continuously, and the natural earthquake can cause the occurrence of the aftershock to be tighter at the moment, and when the aftershock concentration is higher, the degree of tightness of the aftershock in the earthquake event is higher;

the method for acquiring the residual shock intensity comprises the following steps:

taking the moment when the seismograph monitors the earthquake waves for the first time as a first moment and the moment when the seismograph monitors the earthquake waves for the last time as a second moment, and measuring the vibration amplitudes of all the earthquake waves from the first moment value to the second moment;

Marking the time corresponding to the maximum value of the vibration amplitude of the seismic wave as a target time, and marking the target time to a second time as an effective period; when the maximum value of the vibration amplitude of the earthquake wave appears, the maximum vibration amplitude of the earthquake wave at the moment is indicated, and the subsequent earthquake wave can be divided into earthquake waves caused by the aftershock, so that the identification accuracy of the aftershock earthquake wave can be improved;

Equally dividing the effective period into Sub-time periods, respectively countThe number of seismic waves monitored during a sub-period is obtainedThe residual shock value;

Will be The aftershock values are compared with the time length corresponding to the subinterval to obtainSub-densities;

The expression of the sub-density is:

；

In the method, in the process of the invention, Is the firstThe degree of sub-concentration is set to,Is the firstThe residual shock value is a value of the residual shock,The time length corresponds to the sub-period;

Based on Generating the sub-intensity and the aftershock intensity;

The expression of the residual shock concentration is:

；

In the method, in the process of the invention, For the concentration degree of the aftershocks,Is the firstSub-densities;

The primary compression difference value refers to the difference between the number difference value of compression and expansion and the standard difference value of longitudinal wave primary motions monitored by the earthquake stations in different directions around the center of the earthquake focus after an earthquake event occurs, and the longitudinal wave primary motions are divided into two types of outward compression and inward expansion when a natural earthquake occurs, wherein the longitudinal wave primary motions monitored by the earthquake stations in different directions are inconsistent when the natural earthquake occurs, the compression and the expansion are the same, and the larger the primary motion compression difference value is, the larger the number difference value of the compression and the expansion is;

the method for acquiring the initial compression difference value comprises the following steps:

Drawing a circle by taking the center of the seismic source as the center of a circle and taking the lengths of three preset safety distances as the radius to obtain a monitoring circle, drawing a line in the monitoring circle after passing through the center of the circle, and equally dividing the monitoring circle into A plurality of regions; by means of circle drawing and area dividing, the seismic stations in different directions within a reasonable distance from the center of the seismic source can be delineated to form a relatively closed data acquisition range, so that the data acquisition range can be reduced, the acquisition effect is improved, interference caused by data acquired by other seismic stations in remote positions is avoided, and the data acquisition stability is improved; the preset safety distance is the minimum distance between the seismic stations capable of completely monitoring the seismic wave signals, so that the deployment distance between two adjacent seismic stations can be set, and enough seismic stations can be defined in a detection circle; the preset safety distance is obtained by selecting the shortest distance between a plurality of historical seismic stations after effectively monitoring the distance between seismic wave signals;

monitoring by seismometer The initial vibration direction of the seismic waves in all the target seismic stations in the individual areas;

The seismic wave with the initial vibration direction of outward compression is recorded as compression wave, the seismic wave with the initial vibration direction of inward expansion is recorded as expansion wave, and statistics is carried out Total sum of compression waves of individual regionsThe total amount of expansion waves for each region;

Will be Total sum of compression waves of individual regionsThe total amount of the expansion waves of each region is subjected to difference comparison one by one to obtainSub-compression differences;

the expression of the sub-compression difference is:

；

In the method, in the process of the invention, Is the firstThe sub-compressed difference value is used to determine,Is the firstThe total amount of compressional waves in the individual zones,Is the firstThe total amount of expansion waves for each region;

The minimum value of the sub-compression difference value is subjected to difference with a preset standard compression difference value, and an initial compression difference value is obtained; the preset standard compression difference value is a numerical basis for defining the number of the compression and expansion of the seismic waves generated in the natural earthquake and the non-natural earthquake, and the magnitude of the sub-compression difference value can be distinguished, so that the sub-compression difference value is divided into the natural earthquake and the non-natural earthquake; the preset standard compression difference value is obtained by acquiring a number between the compression and expansion amounts of seismic waves generated in the process of historic massive natural earthquakes and non-natural earthquakes and then obtaining an average value of the number;

the expression of the initial compression difference is:

；

In the method, in the process of the invention, For the initial compression difference value,Is the minimum value of the sub-compressed difference value,Is the standard compression difference;

The event type value refers to a specific type of the seismic event, so that the type of the seismic event is identified and judged, and the event type value comprises a natural seismic event and a non-natural seismic event; when the event type value is a natural seismic event, the seismic event at the moment is indicated to be a seismic under a natural environment, and when the event type value is a non-natural seismic event, the seismic event at the moment is indicated to be a seismic under a non-natural environment such as artificial explosion or collapse; the event type value is obtained through corresponding event characteristic data under the natural seismic event and the non-natural seismic event;

S2: training a machine learning model for predicting an event type value based on historical training data;

the training method of the machine learning model for predicting the event type value comprises the following steps:

Converting event feature data into a corresponding group of feature vectors, taking the feature vectors as input of a machine learning model, taking event type values corresponding to each group of event feature data as output of the machine learning model, taking the event type values as prediction targets, taking the sum of prediction errors of all training data to be minimized as a training target, and training the machine learning model until the sum of the prediction errors reaches convergence;

Illustratively, the machine learning model is any one of a CNN neural network model or AlexNet;

The calculation formula of the prediction error is as follows:

；

In the method, in the process of the invention, In order to predict the error of the signal,Group number for feature vector; Is the first Predicted state values corresponding to the set of feature vectors,Is the firstThe actual state value corresponding to the group training data;

In the machine learning model, the feature vector is event feature data, the state value is event type value, other model parameters of the machine learning model, the target loss value, the optimization algorithm, the training set test set verification set proportion, the loss function optimization and the like are realized through actual engineering, and the model is obtained after experimental tuning is continuously carried out;

s3: collecting real-time event characteristic data, and inputting the real-time event characteristic data into a machine learning model after training to predict a real-time event type value;

Collecting real-time event feature data of the seismic event, and outputting an event type value corresponding to the real-time event feature data through a machine learning model;

When the output of the event type value of the machine learning model is 1, the earthquake event is indicated to be an earthquake under a natural environment, and the event type value is a natural earthquake event; when the event type value of the machine learning model is 0, the earthquake event is an earthquake under non-natural environments such as artificial explosion or collapse, and the event type value is a non-natural earthquake event;

S4: collecting negative influence parameters of a natural seismic event, and generating natural ground earthquake shadow loudness based on the negative influence parameters;

when the predicted real-time event type value is a natural earthquake event, the occurrence of the natural earthquake is indicated, at the moment, the natural earthquake can cause certain negative damage to facilities such as buildings, crops and the like, in order to acquire the negative damage degree caused by the natural earthquake, the relevant data of the natural earthquake event need to be acquired, and the negative influence parameters are parameters indicating that the facilities such as the buildings, the crops and the like can be negatively damaged in the natural earthquake event;

Negative impact parameters include seismic magnitude and duration ratio; the seismic magnitude value refers to a value of the Rich magnitude calculated by logarithm of the maximum amplitude value of the seismic waves recorded by a seismograph at a specific distance from the center of the seismic source, when the value of the Rich magnitude is larger, the larger the seismic magnitude is, at the moment, the larger harm is brought to buildings and other facilities on the ground by the earthquake, and the greater the damage degree of the earthquake is, the greater the damage degree of natural earthquake events is;

the method for acquiring the earthquake magnitude value comprises the following steps:

drawing a circle by taking the center of the seismic source as a starting point and taking the length 100 kilometers away from the starting point as a radius to obtain a screening circle;

Marking the earthquake station on the screening circle, and connecting a starting point and a terminal point by taking the position of the center of the urban area as the terminal point to obtain a magnitude line; by generating the magnitude line, the earthquake stations on the screening circle can be selected linearly, a unidirectional selection basis of the earthquake stations is obtained, and the efficient and accurate screening effect of the earthquake stations is realized;

when the earthquake station is arranged at the superposition position of the magnitude line and the screening circle, marking the earthquake station at the superposition position of the magnitude line and the screening circle as an available earthquake station;

When no seismic station exists at the overlapping position of the magnitude line and the screening circle, marking the seismic station which is on the screening circle and has the smallest distance from the overlapping position of the magnitude line and the screening circle as the available seismic station; by selecting the seismic station closest to the superposition position of the magnitude line and the screening circle, the seismic station which does not exist on the magnitude line can be effectively replaced, so that the selected available seismic station is always kept closest to the center of the urban area, and the applicability of the available seismic station is improved;

Obtaining the maximum amplitude of the seismic wave received at 100 km from the center of the seismic source and the maximum amplitude of the seismic wave of the 0-level seismic wave received at 100 km from the center of the seismic source through the available seismic station, and generating a seismic magnitude value;

the expression of the seismic magnitude value is:

；

In the method, in the process of the invention, For the magnitude of the earthquake,For the maximum amplitude of the received seismic waves 100 km from the center of the source,Maximum amplitude of seismic waves for a level 0 seismic received 100 km from the center of the source;

The duration ratio is the ratio of the duration time of the earthquake acceleration reaching the preset acceleration to the total duration time when the natural earthquake occurs, and the greater the duration time ratio is, the greater the duration time of the earthquake acceleration reaching the preset acceleration is, the greater the hazard degree caused by the natural earthquake event is;

The method for acquiring the duration occupying ratio comprises the following steps:

random selection within a screening circle Individual seismic stations and acquiringThe value of the acceleration sensor in each seismic station;

Recording Obtaining the time period from the occurrence of the fluctuation to the no-fluctuation of the value of the acceleration sensor in each seismic stationTotal time length of each acceleration change;

After the value of the acceleration sensor is increased to the lower limit value of the acceleration, the time period corresponding to the time period when the value of the acceleration sensor is reduced to the lower limit value of the acceleration is recorded as an acceleration period, and the acceleration period is obtained A plurality of acceleration periods; the lower limit value of the acceleration refers to the minimum value of the earthquake acceleration when the earthquake acceleration can cause damage to the building, and when the earthquake acceleration is larger than the lower limit value of the acceleration, natural earthquake events can cause damage to the building, so that the lower limit of the earthquake acceleration can be limited; the lower acceleration limit is obtained by acquiring the minimum earthquake acceleration when a great number of natural earthquake events cause damage to the building and then obtaining the average value of the minimum earthquake acceleration; one acceleration period represents the time length of the seismic acceleration falling to the acceleration lower limit value again after the seismic acceleration reaches the acceleration lower limit value and rises, so that one rising and falling change period of the seismic acceleration is marked, and a plurality of accurate acceleration time length data are obtained;

Will be After the acceleration periods are accumulated, the total time length of the acceleration changes is compared to obtainA sub-occupation ratio;

the expression of the sub-ratio is:

；

In the method, in the process of the invention, Is the firstThe ratio of the sub-occupation ratios,Is the firstThe first seismic stationThe time period of the acceleration is a period,Is the firstThe total time length of the acceleration change of each seismic station;

Will be Averaging after accumulating the sub-occupation ratios to obtain a duration occupation ratio;

the duration ratio is expressed as:

；

In the method, in the process of the invention, In order for the duration to be a fraction of the time,Is the firstA sub-occupation ratio;

The natural land earthquake shadow loudness refers to the damage severity of the natural earthquake event to facilities such as buildings on the ground and personal property, so that the damage severity of the natural earthquake event can be numerically represented, and the subsequent evaluation and management of the damage severity of the natural earthquake event are facilitated;

The expression for natural earthquake shadow loudness is:

；

In the method, in the process of the invention, In order to naturally be earthquake shadow loudness degrees,Is a natural constant which is used for the preparation of the composite material,、Is a weight factor;

Wherein, The dimensions of the exemplary,At the level of 0.636 of the total weight of the composition,0.364;

It should be noted that, the size of the weight factor is a specific numerical value obtained by quantizing each data, so that the subsequent comparison is convenient, and the size of the weight factor depends on the number of negative influence parameters and the corresponding weight factor is preliminarily set for each group of negative influence parameters by a person skilled in the art;

s5: comparing the loudness of the natural land earthquake shadow with a preset natural earthquake influence threshold value, dividing earthquake influence levels, and generating earthquake influence early warning prompts;

The earth earthquake shadow sound level is the basis for classifying the damage caused by the natural earthquake event, the natural earthquake event with different damage degrees has different damage severity to facilities such as buildings, and the earthquake damage level needs to be identified in order to accurately identify the damage severity of the natural earthquake event, so that the subsequent early warning and management processing are convenient;

the earthquake influence level comprises a first-level early warning level, a second-level early warning level and a third-level early warning level; the severity of the natural seismic event hazard corresponding to the first-level early warning level is smaller than that of the natural seismic event hazard corresponding to the second-level early warning level, and the severity of the natural seismic event hazard corresponding to the second-level early warning level is smaller than that of the natural seismic event hazard corresponding to the third-level early warning level;

the method for dividing the primary early warning level, the secondary early warning level and the tertiary early warning level comprises the following steps:

Natural seismic influence degree With a preset first natural seismic impact thresholdAnd a second natural seismic impact thresholdIn comparison with the comparison result of the comparison,Less than; The first natural seismic impact threshold and the second natural seismic impact threshold are the basis for classifying the severity of the hazard of natural ground earthquake shadow loudness into light, medium and heavy grades, thereby realizing different grade classification for different natural ground earthquake shadow loudness; the first natural earthquake influence threshold and the second natural earthquake influence threshold are debugged for many times after the corresponding boundary values of the natural land earthquake shadow loudness when the hazard severity of the natural land earthquake shadow loudness is light, medium and heavy in a large amount are collected;

When (when) Less than or equal toWhen the loudness of the natural land earthquake shadow of the natural earthquake event is smaller than or equal to a first natural earthquake influence threshold, and the hazard severity degree between the natural earthquakes is light, the natural land earthquake shadow is divided into first-level early warning levels;

When (when) Less thanAnd (2) andLess than or equal toWhen the loudness of the natural land earthquake shadow of the natural earthquake event is larger than the first natural earthquake influence threshold and smaller than or equal to the second natural earthquake influence threshold, the damage severity between the natural earthquakes is divided into a second-level early warning level;

When (when) Greater thanWhen the loudness of the natural land earthquake shadow of the natural earthquake event is larger than the second natural earthquake influence threshold, the hazard severity degree between the natural earthquakes is heavy, and the natural land earthquake shadow is divided into three levels of early warning;

After the natural earthquake event is subjected to earthquake influence level division, the divided influence level is required to be output and displayed on a visual terminal of the earthquake station, so that a foundation for earthquake early warning is provided for management staff in the earthquake station, and data support is provided for subsequent earthquake resistance and disaster reduction, and whether to generate an earthquake influence early warning prompt is required to be judged according to the influence level;

The method for generating the earthquake influence early warning prompt comprises the following steps:

When the earthquake influence level is a first-level early warning level, the damage severity of the natural earthquake event at the moment can be almost ignored, and then no earthquake influence early warning prompt is generated;

when the earthquake influence level is a secondary early warning level, the natural earthquake event hazard severity is medium at the moment, and then an earthquake influence early warning prompt is generated;

When the earthquake influence level is three-level early warning level, the natural earthquake event has serious hazard severity, and then earthquake influence early warning prompt is generated;

after the earthquake influence early warning prompt is generated, the visual terminal of the earthquake station displays relevant parameters of the natural earthquake event, so that necessary earthquake data are provided for management staff, and the natural earthquake event is conveniently and subsequently early-warned;

In the embodiment, through collecting historical training data of a seismic event, the historical training data comprises event feature data and event type values, a machine learning model for predicting the event type values is trained based on the historical training data, real-time event feature data is collected, real-time event type values are predicted in the trained machine learning model, negative influence parameters of the natural seismic event are collected, natural land earthquake shadow loudness is generated based on the negative influence parameters, natural land earthquake shadow loudness is compared with a preset natural seismic influence threshold value, seismic influence levels are divided, and seismic influence early warning prompts are generated; compared with the prior art, the method has the advantages that the event characteristic data can be accurately acquired, the application range of the prediction recognition data of the earthquake event types is enlarged, the machine learning model can accurately predict the earthquake event types on the basis of diversified data, when the natural earthquake event is recognized, the damage severity of the natural earthquake event is adaptively calculated and evaluated, and the damage early warning prompt of the natural earthquake event is given by combining the evaluation result, so that a good basis is provided for the manager of the earthquake station to quickly, accurately and timely send out the earthquake early warning information.

Example 2: referring to fig. 2, a part of the description of embodiment 1 is not described in detail in this embodiment, and a deep learning-based seismic event type identification and determination system is provided, which is used for implementing a deep learning-based seismic event type identification and determination method, and includes a first data acquisition module, a model training module, a real-time prediction module, a second data acquisition module and an early warning prompt module, where the modules are connected by a wired or wireless network manner;

The first data acquisition module is used for acquiring historical training data of the seismic event, wherein the historical training data comprises event characteristic data and event type values;

the model training module is used for training a machine learning model for predicting the event type value based on the historical training data;

The real-time prediction module is used for collecting real-time event characteristic data, inputting the real-time event characteristic data into the trained machine learning model and predicting a real-time event type value;

The second data acquisition module is used for acquiring negative influence parameters of the natural seismic event and generating natural earthquake shadow loudness based on the negative influence parameters;

And the early warning prompt module is used for comparing the loudness of the natural land earthquake shadow with a preset natural earthquake influence threshold value, dividing the earthquake influence level and generating an earthquake influence early warning prompt.

The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (6)

1. The earthquake event type identification judging method based on deep learning is characterized by comprising the following steps of:

s1: collecting historical training data of the seismic event, wherein the historical training data comprises event characteristic data and event type values;

The event characteristic data comprise an aspect ratio difference value, a frequency vibration value, a residual vibration concentration and an initial compression difference value;

The method for acquiring the aspect ratio difference value comprises the following steps:

screening out with the center of the seismic source as the starting point and the length greater than a preset safety distance as the standard The earthquake stations are marked as target earthquake stations;

Acquisition of The occurrence times of longitudinal waves and transverse waves monitored by each target seismic station in a preset acquisition period are respectively obtainedLongitudinal wave valuesA transverse wave value;

Will be The longitudinal wave values are respectively compared with the time length corresponding to the preset acquisition period to obtainSub-longitudinal wave values;

The expression of the sub-longitudinal wave value is:

；

In the method, in the process of the invention, Is the firstThe number of sub-longitudinal wave values,Is the firstThe number of longitudinal wave-length values,The method comprises the steps of presetting a duration corresponding to a collection period;

Will be The transverse wave values are respectively compared with the time length corresponding to the preset acquisition period to obtainSub-transverse wave values;

The expression of the sub-transverse wave value is:

；

In the method, in the process of the invention, Is the firstThe value of the sub-transverse wave,Is the firstA transverse wave value;

Removing maximum value and minimum value of sub longitudinal wave value and sub transverse wave value respectively, and keeping the rest After the summation of the sub-longitudinal wave values, the sub-longitudinal wave values are combined withComparing the accumulated values of the sub-transverse wave values to obtain a longitudinal-transverse wave ratio;

The expression of the longitudinal-transverse wave ratio is:

；

In the method, in the process of the invention, In order to achieve an aspect ratio of the waves,Is the firstThe number of sub-longitudinal wave values,Is the firstSub-transverse wave values;

After the longitudinal-transverse wave ratio is subjected to difference with the standard ratio, an aspect ratio difference value is obtained;

the expression of the aspect ratio difference is:

；

In the method, in the process of the invention, Is the difference in the aspect ratio,Is a standard ratio;

The method for acquiring the frequency vibration value comprises the following steps:

Measuring the distance between a target seismic station and the center of a seismic source, marking the target seismic station corresponding to the minimum distance value as a seismic station to be detected, and monitoring the seismic wave oscillation frequency of the seismic station to be detected through a seismograph;

The time when the vibration frequency of the earthquake wave reaches the preset lower limit frequency for the first time is recorded as the starting time, and the time when the vibration frequency of the earthquake wave reaches the preset upper limit frequency for the first time is recorded as the ending time;

after the end time and the start time are differenced, an oscillation period is obtained;

The expression of the oscillation period is:

；

In the method, in the process of the invention, For the duration of the oscillation,In order to finish the moment of time,Is the starting time;

Randomly marking in the oscillation period by taking the time length corresponding to one tenth of the oscillation period as a standard Each of the sub-sections not adjacent to each other are countedTotal number of oscillations within the subsection;

Will be The total times of oscillation in the subsections are respectively compared with the time length corresponding to the subsections to obtainSub-frequencies;

the expression of the sub-frequencies is:

；

In the method, in the process of the invention, Is the firstThe sub-frequencies of the frequency band are selected,Is the firstTotal number of oscillations within the subsection;

Will be The sub-frequencies are accumulated and averaged to obtain a frequency vibration value;

the expression of the frequency vibration value is:

；

In the method, in the process of the invention, As the value of the frequency oscillation,Is the firstSub-frequencies;

the method for acquiring the residual shock intensity comprises the following steps:

taking the moment when the seismograph monitors the earthquake waves for the first time as a first moment and the moment when the seismograph monitors the earthquake waves for the last time as a second moment, and measuring the vibration amplitudes of all the earthquake waves from the first moment value to the second moment;

Marking the time corresponding to the maximum value of the vibration amplitude of the seismic wave as a target time, and marking the target time to a second time as an effective period;

Equally dividing the effective period into Sub-time periods, respectively countThe number of seismic waves monitored during a sub-period is obtainedThe residual shock value;

Will be The aftershock values are compared with the time length corresponding to the subinterval to obtainSub-densities;

The expression of the sub-density is:

；

In the method, in the process of the invention, Is the firstThe degree of sub-concentration is set to,Is the firstThe residual shock value is a value of the residual shock,The time length corresponds to the sub-period;

Based on Generating the sub-intensity and the aftershock intensity;

The expression of the residual shock concentration is:

；

In the method, in the process of the invention, For the concentration degree of the aftershocks,Is the firstSub-densities;

the method for acquiring the initial compression difference value comprises the following steps:

Drawing a circle by taking the center of the seismic source as the center of a circle and taking the lengths of three preset safety distances as the radius to obtain a monitoring circle, drawing a line in the monitoring circle after passing through the center of the circle, and equally dividing the monitoring circle into A plurality of regions;

monitoring by seismometer The initial vibration direction of the seismic waves in all the target seismic stations in the individual areas;

The seismic wave with the initial vibration direction of outward compression is recorded as compression wave, the seismic wave with the initial vibration direction of inward expansion is recorded as expansion wave, and statistics is carried out Total sum of compression waves of individual regionsThe total amount of expansion waves for each region;

Will be Total sum of compression waves of individual regionsThe total amount of the expansion waves of each region is subjected to difference comparison one by one to obtainSub-compression differences;

the expression of the sub-compression difference is:

；

In the method, in the process of the invention, Is the firstThe sub-compressed difference value is used to determine,Is the firstThe total amount of compressional waves in the individual zones,Is the firstThe total amount of expansion waves for each region;

the minimum value of the sub-compression difference value is subjected to difference with a preset standard compression difference value, and an initial compression difference value is obtained;

the expression of the initial compression difference is:

；

In the method, in the process of the invention, For the initial compression difference value,Is the minimum value of the sub-compressed difference value,Is the standard compression difference;

S2: training a machine learning model for predicting an event type value based on historical training data;

S3: collecting real-time event characteristic data, inputting the real-time event characteristic data into a machine learning model after training to predict real-time event type values, wherein the event type values comprise natural seismic events and non-natural seismic events;

S4: collecting negative influence parameters of a natural seismic event, and generating natural ground earthquake shadow loudness based on the negative influence parameters;

S5: and comparing the loudness of the natural land earthquake shadow with a preset natural earthquake influence threshold value, dividing earthquake influence levels, and generating earthquake influence early warning prompts.

2. The deep learning based seismic event type identification decision method of claim 1, wherein the training method of the machine learning model for predicting event type values comprises:

Converting the event feature data into a corresponding set of feature vectors;

And taking the feature vector as input of a machine learning model, taking event type values corresponding to each group of event feature data as output of the machine learning model, taking the event type values as prediction targets, taking the sum of prediction errors of all the minimized training data as a training target, and training the machine learning model until the sum of the prediction errors reaches convergence, and stopping training.

3. The depth learning based seismic event type identification decision method of claim 2, wherein when the event type value is a natural seismic event, the output of the machine learning model is 1;

When the event type value is a non-natural seismic event, the output of the machine learning model is 0.

4. The depth learning based seismic event type identification decision method of claim 3 wherein the negative impact parameters include a seismic magnitude value and a duration ratio:

the method for acquiring the earthquake magnitude value comprises the following steps:

drawing a circle by taking the center of the seismic source as a starting point and taking the length 100 kilometers away from the starting point as a radius to obtain a screening circle;

Marking the earthquake station on the screening circle, and connecting a starting point and a terminal point by taking the position of the center of the urban area as the terminal point to obtain a magnitude line;

when the earthquake station is arranged at the superposition position of the magnitude line and the screening circle, marking the earthquake station at the superposition position of the magnitude line and the screening circle as an available earthquake station;

When no seismic station exists at the overlapping position of the magnitude line and the screening circle, marking the seismic station which is on the screening circle and has the smallest distance from the overlapping position of the magnitude line and the screening circle as the available seismic station;

Obtaining the maximum amplitude of the seismic wave received at 100 km from the center of the seismic source and the maximum amplitude of the seismic wave of the 0-level seismic wave received at 100 km from the center of the seismic source through the available seismic station, and generating a seismic magnitude value;

the expression of the seismic magnitude value is:

；

In the method, in the process of the invention, For the magnitude of the earthquake,For the maximum amplitude of the received seismic waves 100 km from the center of the source,Is the maximum amplitude of the seismic wave of a class 0 seismic received 100 km from the center of the source.

5. The method for determining the type of the seismic event based on deep learning according to claim 4, wherein the method for acquiring the duration occupying ratio comprises:

random selection within a screening circle Individual seismic stations and acquiringThe value of the acceleration sensor in each seismic station;

Recording Obtaining the time period from the occurrence of the fluctuation to the no-fluctuation of the value of the acceleration sensor in each seismic stationTotal time length of each acceleration change;

Will be After the acceleration periods are accumulated, the total time length of the acceleration changes is compared to obtainA sub-occupation ratio;

the expression of the sub-ratio is:

；

In the method, in the process of the invention, Is the firstThe ratio of the sub-occupation ratios,Is the firstThe first seismic stationThe time period of the acceleration is a period,Is the firstThe total time length of the acceleration change of each seismic station;

Will be Averaging after accumulating the sub-occupation ratios to obtain a duration occupation ratio;

the duration ratio is expressed as:

；

In the method, in the process of the invention, In order for the duration to be a fraction of the time,Is the firstA sub-occupation ratio;

The expression for natural earthquake shadow loudness is:

；

In the method, in the process of the invention, In order to naturally be earthquake shadow loudness degrees,Is a natural constant which is used for the production of the high-temperature-resistant ceramic material,、Is a weight factor.

6. The depth learning based seismic event type identification determination method of claim 5, wherein the seismic impact level comprises a primary early warning level, a secondary early warning level and a tertiary early warning level;

the method for dividing the primary early warning level, the secondary early warning level and the tertiary early warning level comprises the following steps:

Natural seismic influence degree With a preset first natural seismic impact thresholdAnd a second natural seismic impact thresholdIn comparison with the comparison result of the comparison,Less than；

When (when)Less than or equal toDividing into first-level early warning levels;

When (when) Less thanAnd (2) andLess than or equal toDividing into two-level early warning levels;

When (when) Greater thanDividing into three levels of early warning;

The method for generating the earthquake influence early warning prompt comprises the following steps:

when the earthquake influence level is a first-level early warning level, generating no earthquake influence early warning prompt;

When the earthquake influence level is a secondary early warning level, generating an earthquake influence early warning prompt;

And when the earthquake influence level is three-level early warning level, generating an earthquake influence early warning prompt.