Seismic wave automatic identification method and system based on multi-source sensor

Abstract

The invention discloses a seismic wave automatic identification method and system based on a multi-source sensor, and the method comprises the steps: converting sensor data of different dimensions into a dimensionless numerical range through a data standardization technology for a denoised multi-dimensional data matrix, and forming a standardized data set; if the support vector machine classification algorithm judges that seismic wave signals exist, performing short-time Fourier transform on the feature data set through a time frequency analysis module to obtain frequency distribution characteristics of seismic waves; according to the frequency distribution characteristics, in combination with a preset threshold range, determining the intensity grade of the seismic wave signal, and generating a corresponding signal intensity classification result; aiming at a signal intensity classification result, a real-time monitoring module is linked, the classification result and historical data are compared and analyzed, and a seismic wave dynamic change trend is obtained; and the early warning information generation module is activated through the seismic wave dynamic change trend, and the final early warning level is determined and an alarm signal is output according to the change trend and the intensity level.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of seismic wave recognition, and particularly relates to a seismic wave automatic recognition method and system based on multi-source sensors. BACKGROUND

[0002] As an important field of disaster prevention and mitigation, earthquake monitoring undertakes the key mission of ensuring the safety of life and property, and its research and application has irreplaceable value for social stability and economic development. Accurate and timely identification of seismic waves and analysis of their characteristics are the core of earthquake warning and disaster assessment, and are directly related to the effectiveness of the warning system. However, current seismic wave recognition methods rely on a single type of sensor, such as using only accelerometers or displacement meters for data acquisition. This approach is often disturbed by environmental noise in complex environments, resulting in decreased recognition accuracy. At the same time, in the face of diversified seismic signals, the data of a single sensor is not enough to fully reflect the dynamic characteristics of seismic waves, limiting the real-time performance and reliability of the system.

[0003] Under this background, the field of seismic wave recognition faces significant technical challenges. The first problem is how to effectively integrate data from different types of sensors. Since the data characteristics collected by speed meters, accelerometers and displacement meters are different, the heterogeneity between data makes the fusion process complex and difficult to form a unified analysis framework. This difficulty in data fusion further leads to another key problem, i.e., how to accurately extract the characteristics of seismic waves and achieve automatic recognition based on multi-source data through appropriate algorithms. If the problems of multi-source data fusion and feature extraction cannot be solved, the system will be difficult to maintain high accuracy in a strong noise environment and will not meet the real-time monitoring requirements, thereby affecting the response speed of earthquake warning.

[0004] Therefore, how to build a system that can effectively fuse multi-source sensor data and achieve automatic recognition of seismic wave characteristics through optimized algorithms has become a key problem in improving earthquake monitoring capabilities. SUMMARY

[0005] To solve the above technical problems, the present application provides a seismic wave automatic recognition method and system based on multi-source sensors to solve the problems existing in the prior art.

[0006] In the first aspect, to achieve the above object, the present application provides a seismic wave automatic recognition method based on multi-source sensors, comprising the following steps:

[0007] Synchronously collecting data by speed meters, accelerometers and displacement meters to form an initial multi-dimensional data set;

[0008] Performing time alignment processing on the initial multi-dimensional data set to generate a multi-source data group with a unified time sequence;

[0009] The multi-source data set of the unified time sequence is denoised to generate a denoised multi-dimensional data matrix;

[0010] The multi-dimensional data matrix is subjected to data standardization processing to form a standardized data set;

[0011] The standardized data set is subjected to dimension reduction processing by principal component analysis to extract main feature vectors and obtain a fused feature data set;

[0012] The feature data set is classified by using a support vector machine classification algorithm to determine whether there is a seismic wave signal;

[0013] The feature data set with the seismic wave signal is subjected to short-time Fourier transform to obtain frequency distribution characteristics;

[0014] The intensity level of the seismic wave signal is determined according to the frequency distribution characteristics and a preset threshold range;

[0015] The intensity level is compared and analyzed with historical data to generate a dynamic change trend of the seismic wave;

[0016] The warning level is determined according to the dynamic change trend and an alarm signal is output.

[0017] Optionally, the time alignment processing process includes: adjusting time sequence deviation by using an interpolation method according to the time stamp of each sensor; if the time stamp deviation exceeds a preset threshold, interpolating and filling in missing data points and deleting redundant data points; repairing missing values of the filled or deleted data set by using the mean value of the front and rear data points to fill in the missing part.

[0018] Optionally, the denoising processing process includes: smoothing filtering the multi-source data set of the unified time sequence to eliminate abnormal fluctuations; decomposing the smoothed data into high-frequency sub-signals and low-frequency sub-signals by wavelet transform; suppressing components with amplitudes exceeding a preset threshold in the high-frequency sub-signals; and reconstructing the suppressed high-frequency sub-signals and low-frequency sub-signals into a denoised multi-dimensional data matrix.

[0019] Optionally, the data standardization processing process includes: converting the dimensions of velocity, acceleration and displacement into a unified unit system; mapping sensor data of different numerical ranges to a preset dimensionless interval by linear transformation; and smoothing adjusting data points deviating from the preset fluctuation range based on the trend of adjacent data points.

[0020] Optionally, the process of the principal component analysis comprises: assigning initial weights according to the contribution degrees of the acceleration, displacement and velocity meter data; if the weights exceed a preset threshold range, dynamically adjusting the weight proportions so that the sum is 1; and performing weighted fusion on the standardized data set according to the adjusted weights to generate a reduced dimension feature vector.

[0021] Optionally, the process of the comparison analysis comprises: dividing the current intensity level into multiple signal segments according to a time window; extracting intensity fluctuation features of each signal segment and performing similarity matching with seismic wave patterns in a historical database; identifying abnormal fluctuation segments according to the matching results; and analyzing the dynamic evolution law of the abnormal segments by a trend tracking tool.

[0022] In a second aspect, the present application further provides a seismic wave automatic identification system based on multi-source sensors, which is used to implement a seismic wave automatic identification method based on multi-source sensors, and the system comprises:

[0023] a multi-source sensor data acquisition module, configured to synchronously acquire data of a velocity meter, an accelerometer and a displacement meter, and generate an initial multi-dimensional data set;

[0024] a time series alignment module, configured to perform time alignment processing on the initial multi-dimensional data set, and generate a multi-source data group with uniform time series;

[0025] a data preprocessing module, configured to perform denoising processing on the multi-source data group with uniform time series, and generate a multi-dimensional data matrix after denoising;

[0026] a data standardization module, configured to perform data standardization processing on the multi-dimensional data matrix, and form a standardized data set;

[0027] a data fusion module, configured to perform dimension reduction processing on the standardized data set by principal component analysis, extract a main feature vector, and generate a fused feature data set;

[0028] a pattern recognition module, configured to perform classification on the feature data set by using a support vector machine classification algorithm, and determine whether there is a seismic wave signal;

[0029] a time-frequency analysis module, configured to perform short-time Fourier transform on the feature data set with the seismic wave signal, and acquire frequency distribution characteristics;

[0030] a signal intensity determination module, configured to determine an intensity level of the seismic wave signal according to the frequency distribution characteristics and a preset threshold range;

[0031] a dynamic trend analysis module, configured to perform comparison analysis on the intensity level and historical data, and generate a seismic wave dynamic change trend;

[0032] The early warning information generation module is configured to determine an early warning level according to the dynamic change trend and output an alarm signal.

[0033] In a third aspect, the present application further provides a computer terminal device, comprising:

[0034] one or more processors;

[0035] a memory coupled to the processors, storing one or more programs;

[0036] When the one or more programs are executed by the one or more processors, the one or more processors implement a method for automatically identifying seismic waves based on multi-source sensors.

[0037] In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement a method for automatically identifying seismic waves based on multi-source sensors.

[0038] Compared with the prior art, the present application has the following advantages and technical effects:

[0039] The present application provides a method and system for automatically identifying seismic waves based on multi-source sensors, which synchronously collects data of a velocity meter, an accelerometer and a displacement meter, time-aligns, denoises and standardizes the multi-source data, constructs a data fusion model and extracts main features, identifies seismic wave signals by using a support vector machine algorithm, and further analyzes frequency distribution characteristics and intensity levels of the seismic waves. The present application also analyzes a dynamic change trend of the seismic waves in combination with historical data, and finally determines an early warning level according to the change trend and the intensity level and outputs an alarm signal. The method realizes accurate identification of seismic wave signals and timely early warning, effectively improves the accuracy and real-time performance of seismic monitoring, and provides important technical support for earthquake early warning and disaster prevention and mitigation. BRIEF DESCRIPTION OF DRAWINGS

[0040] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application, and are used to interpret the illustrative embodiments of the present application and their descriptions, and do not constitute improper limitations to the present application. In the drawings:

[0041] Figure 1 is a method flowchart of an embodiment of the present application;

[0042] Figure 2 is a part of the method flowchart of an embodiment of the present application;

[0043] Figure 3 is a part of the method flowchart of an embodiment of the present application. DETAILED DESCRIPTION

[0044] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0045] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described herein can be executed in an order different from that shown.

[0046] Embodiment one

[0047] As shown in Figure 1 , Figure 2 , Figure 3 , the present embodiment provides a method for automatic identification of seismic waves based on multi-source sensors, comprising:

[0048] Synchronously collecting data by a velocity meter, an accelerometer and a displacement meter to form an initial multi-dimensional data set;

[0049] Time alignment processing is performed on the initial multi-dimensional data set to generate a multi-source data group with uniform time sequence;

[0050] The multi-source data group with uniform time sequence is subjected to denoising processing to generate a multi-dimensional data matrix after denoising;

[0051] The multi-dimensional data matrix is subjected to data standardization processing to form a standardized data set;

[0052] The standardized data set is subjected to dimension reduction processing by principal component analysis to extract main feature vectors, thereby obtaining a fused feature data set;

[0053] The feature data set is classified by using a support vector machine classification algorithm to determine whether there is a seismic wave signal;

[0054] Short-time Fourier transform is performed on the feature data set with the seismic wave signal to obtain frequency distribution characteristics;

[0055] According to the frequency distribution characteristics and a preset threshold range, the intensity level of the seismic wave signal is determined;

[0056] The intensity level is compared and analyzed with historical data to generate a dynamic change trend of the seismic wave;

[0057] According to the dynamic change trend, a warning level is determined and an alarm signal is output.

[0058] As an implementation in the embodiment, the process of the time alignment processing includes: adjusting the time sequence deviation by using an interpolation method according to the timestamp identification of each sensor; if the timestamp deviation exceeds a preset threshold, interpolating the missing data points and deleting the redundant data points; repairing the missing values of the data set after the interpolation or deletion, and filling the missing part by using the mean value of the front and rear data points.

[0059] Specifically, the specific process includes:

[0060] In step S101, the data of the speedometer, the accelerometer and the displacement meter are synchronously acquired by the multi-source sensor data acquisition module to form an initial multi-dimensional data set, the timestamps of different sensors are aligned, and a multi-source data set with uniform time sequence is obtained.

[0061] More specifically, the process includes: acquiring the acceleration data, the displacement data and the speed data by using a uniform acquisition frequency through the multi-source sensor module, and attaching a timestamp to the data of each sensor during acquisition to obtain an initial multi-dimensional data set with time identification. According to the initial multi-dimensional data set, the timestamps of different sensors are processed, a time alignment algorithm is used, the time sequence is adjusted by using an interpolation method, it is judged whether the timestamp deviation exceeds a preset threshold, if it exceeds, the data points are interpolated or deleted, and a multi-dimensional data set with consistent time is obtained. From the multi-dimensional data set with consistent time, the data integrity information is acquired, the acceleration data, the displacement data and the speed data are repaired by using the mean value filling method for the missing part that may exist in the data, it is determined that the repaired data meets the consistency requirement, and a complete multi-source data set is obtained. Through the complete multi-source data set, the acceleration data, the displacement data and the speed data are comprehensively processed by using a weighted average method for the sensor fusion requirement, it is judged whether the data weight meets the preset threshold range, if not, the weight proportion is adjusted, and a final unified fusion data set is obtained.

[0062] For example, in the multi-source sensor data processing scene, the speed, acceleration and displacement three kinds of sensor data are collected and fused, aiming to improve the accuracy of device state evaluation. First, for the uniform acquisition frequency and timestamp processing, it is assumed that the acquisition frequency is set to 100 times per second, and each sensor data will automatically attach a timestamp during acquisition to form an initial multi-dimensional data set. Assuming that the accelerometer collects 100 data points in a second, the displacement and speed meter may only collect 98 data points due to hardware delay, and the timestamp will record the specific acquisition time of each data point, such as 2023-10-01 10:00:00.001 to 2023-10-01 10:00:00.999.

[0063] Specifically, for the time alignment algorithm, the time series is adjusted by interpolation method, assuming that the preset timestamp deviation threshold is 5 milliseconds, if the timestamp deviation of a displacement data point is 6 milliseconds, the data point is filled by linear interpolation to ensure alignment with the acceleration data. If the speed data point deviation is too large and cannot be filled by interpolation, the redundant data points are deleted, and finally the time-consistent multi-dimensional data group is obtained. This alignment method can effectively reduce the data analysis error caused by time deviation and improve the subsequent fusion accuracy.

[0064] For example, in data integrity processing, from the time-consistent multi-dimensional data group, it is found that the acceleration data is missing 5 data points in a certain time period, the mean filling method can be used to fill the missing part with the average of the data points before and after. Assuming that the data values before and after the missing point are 2.5 and 3.5, respectively, the filling value is 3.0, and similar methods are also applicable to displacement and speed data repair. After repair, consistency check is performed to ensure that the data fluctuation is within a reasonable range, and a complete multi-source data group is obtained. This method can ensure data continuity and avoid the impact of missing data on subsequent analysis.

[0065] Specifically, for the sensor fusion requirement, a weighted average method is used to comprehensively process multi-source data. Assuming that the initial weights are acceleration 0.4, displacement 0.3, and speed 0.3, if the preset weight threshold range is 0.2 to 0.5, it is found that they all meet the requirements, so there is no need to adjust. If the acceleration weight is calculated as 0.6, which exceeds the threshold, it is adjusted to 0.5, and the other weight proportions are correspondingly increased to ensure the sum is 1. The final fused data group can comprehensively reflect the device state, for example, in the case of vibration anomaly, the fused data can more accurately indicate the fault risk. The weighted average method can balance the contribution of each sensor data and improve the monitoring reliability.

[0066] For example, from data repair to fusion, the whole process avoids data deviation through time alignment, ensures data integrity through mean filling, and optimizes fusion effect through weight adjustment. These steps together ensure the high quality of the final data group, provide a reliable basis for industrial equipment state monitoring, significantly reduce the misjudgment rate, and improve the system decision efficiency.

[0067] As an embodiment in this embodiment, the process of the denoising processing includes: smoothing filtering the multi-source data group of the unified time series to eliminate abnormal fluctuations; decomposing the smoothed data into high-frequency sub-signals and low-frequency sub-signals through wavelet transform; suppressing components with amplitudes exceeding a preset threshold in the high-frequency sub-signals; and reconstructing the suppressed high-frequency sub-signals and low-frequency sub-signals into a denoised multi-dimensional data matrix.

[0068] Specifically, the specific process includes:

[0069] In step S102, according to the unified time sequence multi-source data set, the data preprocessing module is used to denoise the sensor data, and the wavelet transform method is used to separate the noise component to generate a denoised multi-dimensional data matrix.

[0070] More specific processes include: according to the multi-source data, the original record of the sensor data is obtained from the time sequence, the pre-established smoothing filter tool is used to preliminarily process the abnormal fluctuations that may exist in the original record to obtain a smoothed basic data set. Through the smoothed basic data set, the wavelet transform tool is used to signal decompose the basic data set for the residual noise component therein, separate the data into sub-signals of different frequencies, and determine the separated sub-signal set. If some frequency components in the sub-signal set exceed a preset threshold, a filter is used to suppress the components exceeding the threshold to obtain a suppressed sub-signal combination. According to the suppressed sub-signal combination, the signal reconstruction tool is used to integrate the sub-signal combination to generate a denoised multi-dimensional data matrix, and it is judged whether the matrix meets the data consistency requirement to obtain a final processing result.

[0071] For example, for the processing of multi-source sensor data, abnormal fluctuations caused by device vibration or environmental interference may exist in the original record. For this problem, the original time sequence data of the acceleration, displacement and speed meter can be preliminarily processed by the smoothing filter tool. Assuming that the acceleration data has a sudden high value fluctuation in a certain period of time, such as 2.5, 8.0, 2.8 for three consecutive data points, the mean smoothing method can be used to adjust the middle abnormal value 8.0 to a value closer to the front and rear data points, such as about 3.0, to obtain a smoothed basic data set. This way can effectively weaken the influence of sudden fluctuations on subsequent analysis.

[0072] For example, for the noise component that may still remain after smoothing, the wavelet transform tool can be used to signal decompose the basic data set. In principle, wavelet transform can decompose the signal into sub-signals of different frequencies, which is convenient for identifying the frequency range of noise. Assuming that the displacement sensor data is decomposed to obtain high-frequency sub-signals and low-frequency sub-signals, and the high-frequency part may contain noise. A preset threshold is set, such as sub-signals with an amplitude exceeding 0.5 are considered abnormal, and if the amplitude of a certain high-frequency sub-signal is 0.7, a filter is used to suppress this part to reduce its amplitude to a reasonable range, such as below 0.3, to obtain a suppressed sub-signal combination. This decomposition and suppression method helps to retain the main characteristics of the signal while reducing noise interference.

[0073] For example, in the signal reconstruction stage, the inhibited sub-signals are combined and integrated into a multi-dimensional data matrix after denoising by the signal reconstruction tool. Assuming that the speedometer data is reconstructed to obtain a multi-dimensional matrix containing time series, it is necessary to further determine whether it meets the data consistency requirements. The fluctuation range between data points can be checked to see if it is within the preset range, such as the fluctuation range of acceleration data should be between 0.1 and 5.0. If the fluctuation of some data exceeds this range, it needs to be re-adjusted or marked as abnormal. This consistency check can ensure the reliability of the final data and provide a solid foundation for subsequent equipment state analysis.

[0074] For example, from a business perspective, the above processing flow is closely related to the monitoring of industrial equipment seismic needs. Abnormal fluctuations in raw data may be caused by sudden vibrations in equipment operation or sensor errors, which can affect the accuracy of state evaluation if not handled in a timely manner. Through steps such as smoothing filtering, wavelet transform and signal reconstruction, the data quality is gradually optimized to ensure the coordination of multi-source data in time series. This method is particularly suitable for efficient processing of multi-sensor data in complex industrial environments, providing important support for real-time monitoring of equipment operating conditions.

[0075] As an embodiment in this embodiment, the process of data standardization processing includes: converting the dimensions of speed, acceleration and displacement into a unified unit system; mapping sensor data with different numerical ranges to a preset dimensionless interval through linear transformation; for data points that deviate from the preset fluctuation range after mapping, smooth adjustment based on the trend of adjacent data points.

[0076] Specifically, the specific process includes:

[0077] Step S103, for the multi-dimensional data matrix after denoising, apply data standardization technology to convert sensor data of different dimensions into a dimensionless numerical range to form a standardized data set.

[0078] More specifically, the process includes:

[0079] According to the denoised multi-dimensional data matrix, the data records of different sensor types are obtained, and for the dimensional differences in the data records, a pre-established dimension conversion tool is used for preliminary processing of each dimension to obtain an initial data set with unified dimensions. Through the initial data set, for the numerical range and data scale in the data set, a standardization calculation tool is used to perform linear transformation on the data, map the values of different ranges to a preset unified interval, and determine the intermediate data combination after mapping. According to the intermediate data combination, for the data distribution and processing accuracy in the intermediate data combination, if some data dimensions in the combination deviate from the preset threshold, a smoothing adjustment tool is used to correct the deviated part to obtain the adjusted data set after correction. Through the adjusted data set, for the data consistency and output format in the adjusted data set, a format conversion tool is used to finally integrate the data to generate a final data set that meets the standardization rules.

[0080] For example, when subsequent processing is performed on the denoised multi-dimensional data matrix, attention needs to be paid to the dimensional differences and numerical ranges of different sensor types. For the dimensional differences, a pre-established dimension conversion tool can be used to convert data of different dimensions such as acceleration, displacement, and speed into a unified unit system. Assuming that acceleration data is in units of meters per second squared and displacement data is in units of millimeters, the displacement data can be converted to meters by the conversion tool for subsequent processing. This conversion ensures the comparability of data in physical meaning.

[0081] For example, for the processing of numerical range and data scale, a standardization calculation tool can map values of different ranges to a preset interval, such as between 0 and 1. Assuming that the original range of acceleration data is between -10 and 10, and the range of speed data is between 0 and 100, after linear transformation, both are mapped to the interval of 0 to 1. This way makes the data of different sensors consistent in numerical scale, laying a foundation for subsequent analysis.

[0082] For example, when processing the intermediate data combination, if some data dimensions deviate from the preset threshold, a smoothing adjustment tool can be used for correction. Assuming that in the value distribution of displacement data in a certain period of time, some data points suddenly deviate and exceed the preset threshold range, such as the normal range of 0.2 to 0.8, and a point value of 1.2, the value is adjusted downward to about 0.7 by the smoothing adjustment tool. This correction avoids the influence of abnormal fluctuations in data distribution on overall analysis, ensuring the stability of data.

[0083] For example, for the final data consistency and output format, the format conversion tool can integrate the adjusted data into a set that conforms to standardized rules. Assuming that the time series format in the adjusted data set is not uniform, with some data in seconds and some in milliseconds, the format conversion tool can unify it to milliseconds and output it according to the preset table structure. This integration ensures the consistency of the data in form, facilitating subsequent system reading and processing. Through the above processing, from dimension conversion to format integration, each link closely revolves around the needs of industrial equipment monitoring. Dimension unification avoids unit conflicts when comparing data, standardized mapping reduces numerical differences, smoothing adjustment reduces abnormal interference, and format conversion improves data usability. These steps collectively ensure the rigor of data processing and provide reliable support for accurate assessment of equipment status.

[0084] As an embodiment in this embodiment, the process of principal component analysis includes: assigning initial weights according to the contribution of acceleration, displacement and speedometer data; if the weight exceeds the preset threshold range, dynamically adjusting the weight proportion so that the sum is 1; weighting and fusing the standardized data set according to the adjusted weight to generate the reduced dimension feature vector.

[0085] As an additional embodiment in this embodiment, the covariance matrix of the standardized data set is calculated by principal component analysis to solve its eigenvalues and eigenvectors; according to the size of the eigenvalues, the eigenvectors corresponding to the first K largest eigenvalues are selected as the principal component directions; the original data is projected onto these principal component directions to obtain the reduced dimension feature data set (principal component score).

[0086] Specifically, the specific process includes:

[0087] Step S104, constructing a multi-source data fusion model through a standardized data set, using principal component analysis method to reduce the dimension of multi-dimensional data, extracting the main feature vector, and determining the fused feature data set.

[0088] More specifically, the process includes:

[0089] According to the standardized data set, the pre-established data integration tool is used to preliminarily merge the set for the multi-source data characteristics in the set, to process the consistency problem in the data set, and to obtain an integrated unified data group. Through the unified data group, the feature extraction is performed on the data group by using the dimension reduction processing tool for the multi-dimensional characteristics in the unified data group, to determine a simplified feature data group. The feature data group is obtained, and for the distribution characteristics in the feature data group, if the distribution of the feature data group deviates from a preset threshold, a smoothing adjustment tool is used to correct the deviated part, to obtain an adjusted balanced data group. According to the balanced data group, the format conversion tool is used to finally arrange the data group for the integration requirement in the balanced data group, to determine a feature data set meeting the fusion target.

[0090] For example, for the multi-source data characteristics of the standardized data set, the pre-established data integration tool can be used for preliminary merging to process the data consistency problem. The core of the data integration tool is to identify and unify the sampling frequency and timestamp difference of data collected by different sensors. Assuming that the data of the accelerometer is sampled at 10 times per second, and the data of the speed meter is sampled at 5 times per second, the integration tool will fill in the speed data to 10 times per second by the cubic spline interpolation method, ensure time alignment, form a unified integrated data group, and eliminate analysis errors caused by sampling rate difference.

[0091] For example, for the multi-dimensional characteristics of the integrated unified data group, the dimension reduction processing tool can be used for feature extraction to simplify the data dimension. The core of the dimension reduction processing is to eliminate the redundant information in the high-dimensional data and retain the key features. Assuming that the original data group contains 30 variables in three dimensions of vibration acceleration, displacement and speed, after analysis by the dimension reduction tool, it is found that 20 variables are highly correlated, and finally 10 main features are extracted to form a simplified feature data group. This way can effectively reduce the data amount for subsequent processing.

[0092] For example, for the distribution characteristics of the feature data group, if it is found that part of the data deviates from the preset threshold, the smoothing adjustment tool needs to be used for correction. Assuming that the normal distribution range of the vibration acceleration in the feature data group is 0.1 to 0.9, but the data value at some time point reaches 1.5, which is obviously beyond the threshold. The smoothing adjustment tool will adjust the abnormal value to about 0.8 according to the data trend of the previous and subsequent time points, to form a balanced data group. This adjustment helps to maintain the continuity of the data distribution.

[0093] For example, to meet the integration requirements of the balanced data set, the format conversion tool can be used for final arrangement to ensure that the data meets the fusion target. Assuming that part of the data in the balanced data set is in the format of hours-minutes, and another part is in the total number of seconds, the format conversion tool will unify all the time into the total number of seconds format and arrange it according to the preset structure, and output the feature data set that meets the requirements. This unified format processing facilitates subsequent system reading and analysis. Through the above processing, from data integration to format conversion, each step closely surrounds the needs of industrial equipment monitoring to ensure the integrity of the data from multiple sources to fusion. The integration tool solves the inconsistency of multi-source data, the dimension reduction tool simplifies the data dimension, the smoothing adjustment tool corrects the abnormal distribution, and the format conversion tool guarantees the standardization of the final output. These steps together provide a solid data foundation for equipment state analysis.

[0094] As an embodiment in this embodiment, in step S105, according to the fused feature data set, a support vector machine classification algorithm is used to perform pattern recognition on the seismic wave signal, and a pre-established training model is used to classify the feature vector to determine whether the seismic wave signal exists.

[0095] Specifically, the specific process includes:

[0096] According to the fused feature data set, a pre-established filtering tool is used to perform preliminary processing on the data set according to the waveform features and data distribution in the data set, and a cleaned balanced data set is obtained. Through the balanced data set, a signal extraction tool is used to separate the balanced data set according to the signal intensity and signal source in the balanced data set, and a layered signal set is determined. The signal set is obtained, and a smoothing adjustment tool is used to correct the signal segment in the signal set if the intensity of the signal segment deviates from a preset threshold, and an adjusted signal data set is obtained. According to the adjusted signal data set, a pattern matching tool is used to process the adjusted signal data set according to the feature vector and classification judgment in the adjusted signal data set, and it is determined whether there is a feature segment that meets the seismic wave signal.

[0097] For example, the preliminary processing of the filtering tool is particularly important for the waveform features and data distribution of the fused feature data set. The main function of the filtering tool is to remove noise interference in the data and retain key signal features. Assuming that the vibration waveform in the data set is affected by background noise and some frequency band signals are hidden, the filtering tool can remove irrelevant low-frequency or high-frequency noise by setting the frequency range to obtain clearer waveform data. For example, the original data may contain invalid interference below 0.5 Hz, and after filtering by the filtering tool, the waveform features are more prominent, laying a foundation for subsequent analysis.

[0098] For example, the hierarchical separation process of the signal extraction tool aims to distinguish signals of different origins in terms of signal strength and signal source in the balanced data set.

[0099] In one possible implementation, assuming that the signals collected in industrial equipment monitoring include vibration signals and environmental interference signals, the signal extraction tool can separate the vibration signals from the mixed data by intensity threshold and time domain features. The specific separation method can be based on the distribution interval of signal strength, such as the vibration signal strength is usually in the range of 1.0 to 20.0, while the environmental interference signal strength is mostly between 0.05 to 1.0, by setting hierarchical standards, the signal set is divided into two parts of main signals and secondary signals. This hierarchical helps to focus more accurately on the core signals related to equipment operation.

[0100] For example, for the seismic wave signal and signal pattern in the signal set, the modification process of the smoothing adjustment tool mainly targets signal segments with abnormal intensity. Assuming that the intensity value of a certain signal segment reaches 8.0, while the preset upper threshold is 6.0, which is obviously deviated from the normal range, the smoothing adjustment tool will adjust the intensity value of this segment to about 5.8 in combination with the signal trend of the previous and next time points. This adjustment method avoids the interference of data mutation on subsequent analysis, while preserving the overall trend characteristics of the signal, ensuring that the adjusted signal data set is more consistent with the actual equipment operation state.

[0101] For example, in the adjusted signal data set, the processing process of the pattern matching tool is used to identify whether there are feature segments that match the characteristics of the seismic wave signal.

[0102] Specifically, the pattern matching tool will load the known seismic wave signal template in advance, and then compare the feature vectors in the data set with the template. Assuming that the feature vector of the seismic wave signal in the template contains specific frequency peaks and waveform periods, the tool will compare the signal segments in the data set one by one, and if it finds that the frequency peak and period of a certain signal segment are highly consistent with the template, it will be judged as a potential seismic wave signal. This way can effectively filter out key signal segments, providing important basis for abnormal early warning of equipment state.

[0103] In one possible implementation, the processing of each link is closely combined with the needs of industrial equipment monitoring, from filtering to pattern matching, each step aims to improve data quality and analysis accuracy. The filtering tool cleans up noise, the signal extraction tool separates signals hierarchically, the smoothing adjustment tool corrects abnormalities, and the pattern matching tool identifies characteristic signals. These links are closely linked to ensure the reliability of the final analysis results, providing solid support for real-time monitoring of equipment operation state and timely discovery of potential risks.

[0104] As an implementation in this embodiment, in step S106, if the support vector machine classification algorithm determines that there is a seismic wave signal, the time-frequency analysis module is used to perform short-time Fourier transform on the feature data set to obtain the frequency distribution characteristics of the seismic wave.

[0105] Specifically, the specific process includes:

[0106] According to the feature data set, the waveform characteristics and data distribution thereof are processed by using a pre-established filtering tool to obtain a cleaned balanced data set. By using a signal extraction tool, the signal intensity and signal segments in the balanced data set are separated in layers to determine a layered signal group. By obtaining the signal group, if the intensity of a signal segment in the signal group exceeds a preset threshold, a smoothing correction tool is used to adjust the signal segment to obtain a corrected signal combination. According to the corrected signal combination, the time analysis and frequency distribution thereof are processed by using a short-time Fourier transform tool to determine whether there is a frequency characteristic segment that meets the seismic wave signal.

[0107] For example, in the field of seismic wave signal monitoring, the preliminary processing of the filtering tool for the waveform characteristics and data distribution of the feature data set is particularly crucial. The core of the filtering tool is to eliminate irrelevant interference in the data through a specific frequency screening mechanism. Assuming that in a monitoring scenario, the original data contains a large amount of low-frequency noise with a frequency lower than 0.5 Hz, while the typical frequency range of the seismic wave signal is between 0.1 and 10.0 Hz, the filtering tool can set a frequency threshold to filter out the part below 0.5 Hz, thereby retaining the key signal characteristics. This processing method can provide a clearer data basis for subsequent analysis.

[0108] For example, the signal extraction tool separates the signal intensity and signal segments in the balanced data set in layers to distinguish signals with different characteristics.

[0109] Specifically, in the data collected by the monitoring device, there may be device vibration signals and external environmental interference signals. The signal extraction tool can analyze the distribution range of signal intensity, for example, the vibration signal intensity is usually between 2.0 and 4.0, while the interference signal is usually between 0.1 and 0.8, and set the corresponding intensity threshold to divide the signals into two groups of main signals and secondary signals. This layered separation helps to focus on the core signal and improve the relevance of subsequent analysis.

[0110] For example, for the seismic wave signal and frequency characteristics in the signal packet, the adjustment process of the smoothing correction tool mainly targets the signal segments with abnormal intensity. Assuming that the intensity value of a certain signal segment reaches 7.5, while the preset upper threshold is 5.0, which is obviously beyond the normal range, the smoothing correction tool will adjust the intensity value of this segment to about 5.2 in combination with the signal trend at the previous and subsequent time points. This adjustment avoids the interference of data mutation on analysis, while retaining the overall characteristics of the signal, laying a foundation for subsequent processing.

[0111] For example, in the corrected signal combination, the processing process of the short-time Fourier transform tool is used to determine whether there is a frequency characteristic segment that meets the seismic wave signal.

[0112] Specifically, the tool can decompose the signal into multiple time windows and analyze the frequency distribution in each window. Assuming that the typical frequency peak of the seismic wave signal is between 0.1 and 10.0 Hz, the tool will check the frequency characteristics of each time window one by one, and if it finds that the frequency peak of a certain signal segment falls within this range, it will determine it as a potential seismic wave signal segment. This method can effectively capture the dynamic change characteristics of the signal and provide a reliable basis for signal recognition.

[0113] As an embodiment in this embodiment, in step S107, according to the frequency distribution characteristics, in combination with the preset threshold range, the intensity level of the seismic wave signal is determined, and the corresponding signal intensity classification result is generated.

[0114] Specifically, the specific process includes:

[0115] According to the frequency distribution characteristics, for the signal segments and waveform characteristics therein, a preset signal filtering tool is used to preliminarily process the collected data to obtain a processed signal set. Through the processed signal set, for the time window and feature extraction therein, a short-time Fourier transform tool is used to segment the signal set for processing, and a segmented frequency feature combination is determined. The frequency feature combination is obtained, for the signal intensity and threshold range therein, if the intensity of a certain frequency feature segment exceeds the preset threshold range, a smoothing adjustment tool is used to correct the frequency feature segment to obtain an adjusted feature data set. According to the adjusted feature data set, for the intensity level and determination standard therein, a pre-established classification tool is used to process the feature data set for classification to determine the corresponding signal intensity classification result.

[0116] For example, in the field of seismic wave signal monitoring, the preliminary processing of signal filtering tools is particularly important for frequency distribution characteristics and signal segment processing. The main function of signal filtering tools is to remove irrelevant interference in collected data through specific frequency screening mechanisms, providing a more pure data basis for subsequent analysis. Assuming that in a monitoring scenario, a large amount of background noise is mixed in the original data, with a frequency range of 0.05 to 0.1 Hz low-frequency pulsation, and the target seismic wave signal frequency is concentrated between 0.1 to 10.0 Hz. The signal filtering tool can set a frequency threshold of 0.1 Hz, and remove noise below this range, thus retaining the key signal characteristics. This processing method can significantly improve the clarity of the data, laying the foundation for subsequent feature extraction.

[0117] For example, for the processed signal set, the segmentation processing of the short-time Fourier transform tool is a key step in determining the frequency feature combination. This tool analyzes the frequency distribution in each window by dividing the signal into multiple time windows to capture the dynamic changes of the signal. Assuming that the signal is divided into 0.5-second time windows, and the frequency peak is detected to be concentrated in the range of 2.2 to 2.8 Hz in a certain window, and this range meets the typical characteristics of seismic waves, then the signal segment in this window will be marked as a potential target signal. Through this segmentation processing, the key parts of the signal can be more accurately located, providing a reliable basis for subsequent intensity analysis.

[0118] For example, in the processing of frequency feature combination, if the signal intensity of a certain frequency feature segment exceeds the preset threshold range, the correction of the smoothing adjustment tool is particularly necessary. Assuming that the upper limit of the preset intensity threshold is 6.0, and the intensity value of a certain segment reaches 8.5, which is obviously out of range. The smoothing adjustment tool will adjust the intensity value of this segment to about 6.3 by combining the intensity trend of adjacent time windows, to avoid the interference of data mutation on subsequent analysis. This correction method can maintain the overall continuity of the signal, while reducing the influence of outliers, providing a more stable basis for further processing of the feature data set.

[0119] For example, for the adjusted feature data set, the classification tool is used for classification of signal intensity. This process relies on pre-established intensity level standards and judgment rules. Assuming that the intensity levels are divided into low, medium and high, corresponding to the ranges of 1.0 to 3.0, 3.1 to 5.0 and 5.1 to 7.0 respectively. In a specific monitoring case, the intensity value of a certain signal segment is 4.2, and the classification tool will classify it as medium intensity, and further confirm whether it is a seismic wave signal by combining other features. This hierarchical processing helps to quickly filter out signal segments of different importance levels, providing classification basis for subsequent in-depth analysis, and improving monitoring efficiency.

[0120] As an implementation in the embodiment, the process of the comparison analysis includes: dividing the current intensity level into multiple signal segments according to a time window; extracting the intensity fluctuation features of each signal segment, and performing similarity matching with the seismic wave patterns in the historical database; identifying abnormal fluctuation segments according to the matching results, and analyzing the dynamic evolution law of the abnormal segments by a trend tracking tool.

[0121] Specifically, the specific process includes:

[0122] Step S108, for the signal intensity classification result, the linkage real-time monitoring module compares and analyzes the classification result with the historical data to obtain the dynamic change trend of the seismic wave.

[0123] More specific process includes:

[0124] Obtain the signal intensity classification result, for the intensity level and waveform features therein, use a preset time window tool to segment and extract the classification result, and obtain a segmented intensity segment set therefrom. According to the segmented intensity segment set, for the intensity fluctuation and historical data therein, use a data comparison tool to match the intensity segment with the historical data archives established in advance one by one to determine the matched fluctuation comparison data set. Through the fluctuation comparison data set, for the dynamic change and change law therein, if the intensity fluctuation in a certain time window exceeds the preset threshold range, use a smoothing processing tool to adjust the data set, and judge the adjusted trend feature combination. Obtain the adjusted trend feature combination, for the trend judgment and monitoring period therein, use a preset trend tracking tool to continuously track the feature combination, and obtain the final output result of the dynamic change trend of the seismic wave.

[0125] For example, in the field of seismic wave signal monitoring, the segmentation and extraction of the signal intensity classification result can be realized by a preset time window tool to realize the fine processing of data. The core of the time window tool is to divide the continuous signal data according to fixed or dynamic time intervals, so as to facilitate subsequent analysis. For example, in a monitoring scene, assuming that the signal data is collected every minute, the time window tool can be set to every 10 seconds, and the data of one minute is divided into 6 segments, and the intensity level and waveform features of each segment are extracted. This way can help identify the change of signal intensity in a short time, and provide basic data for subsequent fluctuation analysis.

[0126] For example, the application of data comparison tools is particularly crucial for comparing the segmented intensity fragments with historical data. Data comparison tools match the intensity fluctuations of the current fragment with records in the historical data archive to find similar patterns or abnormal points. Assuming that the historical data archive stores the intensity fluctuation records of seismic waves in a certain region over the past year, and the current fragment shows that the intensity value rapidly rises from 3.5 to 5.2 in a 10-second window, and there is a similar fluctuation pattern in the historical data that is often associated with small earthquakes, then the comparison tool will mark this fragment as a high-concern fragment. This matching process helps quickly locate potential risk signals and improves the targeting of monitoring.

[0127] For example, in the processing of fluctuation comparison data sets, the intervention of smoothing tools is essential if the intensity fluctuation in a certain time window exceeds the preset threshold range. The role of smoothing tools is to reduce the mutation points in the data and maintain the continuity of the trend. Assuming that the upper limit of the preset intensity fluctuation threshold is 1.5, and the intensity value in a window rises from 4.0 to 6.0, which is obviously beyond the range, the smoothing tool will adjust the mutation point to about 5.0 by referring to the intensity values of the previous and next windows. This adjustment can avoid misjudgment caused by a single abnormal value, while preserving the overall trend characteristics of the data, providing a more reliable basis for subsequent analysis.

[0128] For example, the continuous tracking of trend tracking tools is a key link for obtaining the dynamic change trend of seismic waves for the combination of adjusted trend characteristics. Trend tracking tools continuously analyze the intensity data of multiple time windows to identify long-term patterns of signal changes. Assuming that in a continuous one-hour monitoring, the intensity value gradually rises from 3.0 to 5.5, and shows periodic fluctuations, the trend tracking tool will record this change rule and judge whether it is a precursor signal of seismic activity in combination with historical data. This continuous tracking method can capture the dynamic evolution of the signal, provide an important reference for early warning, and significantly improve the timeliness of monitoring.

[0129] As an embodiment in this embodiment, in step S109, the pre-warning information generation module is activated by the dynamic change trend of seismic waves, and the final warning level is determined according to the change trend and intensity level, and an alarm signal is output.

[0130] Specifically, the process includes:

[0131] The trend data of the seismic waveform is acquired, the segmented analysis of the trend data is performed on the fluctuation characteristics and real-time data by using a preset fluctuation comparison tool, and a segmented fluctuation characteristic set is obtained. According to the segmented fluctuation characteristic set, the fluctuation characteristic set is matched with a pre-established intensity standard archive according to the intensity level and level division, and a corresponding intensity level combination is determined. Through the intensity level combination, the intensity level combination is graded by using a level evaluation tool if the intensity level combination exceeds a preset threshold range, and an adaptive warning level result is determined. The warning level result is acquired, and the warning level result is converted into a corresponding alarm signal form by using a signal conversion tool according to the alarm signal and signal output, and final alarm signal data is obtained.

[0132] For example, in the field of seismic wave monitoring, the fluctuation characteristics can be deeply analyzed by various tools and methods for the acquisition and analysis of trend data. The core of the application of the fluctuation comparison tool is to segment the continuous seismic wave data according to time or intensity characteristics, so as to capture the change rule of the signal more carefully. Assuming that the system collects data every minute in a monitoring scene, the fluctuation comparison tool can divide the data into 10-second units, obtaining 6 fluctuation segments, each of which records the peak and valley characteristics in the time period. This segmentation method helps to identify abnormal fluctuations in a short time and lays a foundation for subsequent feature extraction.

[0133] For example, for the matching of the segmented fluctuation characteristic set and the intensity standard archive, the role of the data mapping tool is to associate the current data with the historical standard. Assuming that the intensity standard archive defines intensity level division from 1.0 to 6.0, corresponding to different risk levels, the data mapping tool will map the average intensity value 3.8 of a certain segment to the corresponding level 3, indicating a medium risk. Through this matching, the intensity level combination of each segment can be quickly determined to provide a basis for subsequent warning.

[0134] For example, in the case where the intensity level combination exceeds the preset threshold range, the application of the level evaluation tool is particularly important. Assuming that the preset threshold range is intensity level 4.0, if the level combination of a certain time period reaches 4.5, the level evaluation tool will divide it into a secondary warning level according to the fluctuation duration and historical relevance. This grading process can help monitoring personnel to more accurately judge the risk level and take timely measures.

[0135] For example, for the conversion of early warning level results into alarm signal forms, the implementation method of the signal conversion tool is to convert abstract level data into intuitive alarm forms. Assuming that a certain early warning level result is a secondary early warning, the signal conversion tool will convert it into a specific sound frequency and flashing light signal to facilitate the rapid identification and response of the monitoring center. This conversion method ensures effective information transmission and improves the speed and accuracy of emergency response.

[0136] It should be noted that the cooperative application of the above-mentioned various tools can form a complete data processing chain in seismic wave monitoring, from the analysis of wave characteristics to the output of the final alarm signal, each step being closely connected. The segmented analysis of wave characteristics can capture signal changes in more detail, accurate risk level positioning can be achieved through data mapping and level evaluation, and timely information transmission can be ensured through signal conversion. The combination of these links not only improves the comprehensiveness of monitoring, but also provides reliable support for earthquake early warning.

[0137] Based on this, the embodiment of the present application provides an automatic seismic wave recognition method based on multiple source sensors. By synchronously collecting velocity meter, accelerometer and displacement meter data, the multiple source data are time-aligned, denoised and standardized, a data fusion model is constructed and main features are extracted, a support vector machine algorithm is used to recognize seismic wave signals, and then the frequency distribution characteristics and intensity level of the seismic wave signals are analyzed. The present application also analyzes the dynamic change trend of the seismic wave by combining historical data, and finally determines the early warning level and outputs the alarm signal according to the change trend and intensity level. The method realizes accurate recognition and timely warning of seismic wave signals, effectively improves the accuracy and real-time performance of seismic monitoring, and provides important technical support for earthquake early warning and disaster prevention and reduction.

[0138] Embodiment two

[0139] In this embodiment, a computer terminal device is provided, comprising:

[0140] one or more processors;

[0141] a memory coupled to the processor, for storing one or more programs;

[0142] When the one or more programs are executed by the one or more processors, the one or more processors implement the method in the above embodiments.

[0143] In this embodiment, a computer readable storage medium having a computer program stored thereon is also provided, and the computer program is executed by a processor to implement the method in the above embodiments.

[0144] In the embodiment, an electronic device is also provided, comprising a memory and a processor, the memory stores a computer program, and the processor is configured to run the computer program to execute the method in the above embodiment.

[0145] The above program can be run in a processor, or can also be stored in a memory (or called computer readable medium), the computer readable medium includes permanent and non-permanent, removable and non-removable media, which can realize information storage by any method or technology. The information can be computer readable instructions, data structure, program module or other data. Examples of computer storage medium include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage device or any other non-transmission medium that can be used to store information accessible by a computing device.

[0146] These computer programs can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the flow Figure 1 One flow or multiple flows and / or blocks Figure 1 One block or multiple blocks, and the corresponding different steps can be realized by different modules.

[0147] In the embodiment, such a device or system is provided. The system is called multi-source sensor based automatic seismic wave identification system, which comprises:

[0148] Multi-source sensor data acquisition module, for synchronously acquiring data of velocity meter, accelerometer and displacement meter, and generating initial multi-dimensional data set;

[0149] Time sequence alignment module, for performing time alignment processing on the initial multi-dimensional data set, and generating multi-source data group with uniform time sequence;

[0150] Data preprocessing module, for performing denoising processing on the multi-source data group with uniform time sequence, and generating multi-dimensional data matrix after denoising;

[0151] Data standardization module, for performing data standardization processing on the multi-dimensional data matrix, and forming standardized data set;

[0152] a data fusion module configured to perform dimension reduction processing on the standardized data set by principal component analysis, extract a main feature vector, and generate a fused feature data set;

[0153] a pattern recognition module configured to classify the feature data set by using a support vector machine classification algorithm, and determine whether an earthquake wave signal exists;

[0154] a time-frequency analysis module configured to perform short-time Fourier transform on the feature data set in which the earthquake wave signal exists, and acquire frequency distribution characteristics;

[0155] a signal strength determination module configured to determine a strength level of the earthquake wave signal according to the frequency distribution characteristics and a preset threshold range;

[0156] a dynamic trend analysis module configured to compare and analyze the strength level with historical data, and generate a dynamic change trend of the earthquake wave;

[0157] a warning information generation module configured to determine a warning level according to the dynamic change trend and output an alarm signal.

[0158] As an implementation manner in the embodiment, the time series alignment module comprises:

[0159] a timestamp processing unit configured to attach a unified time identifier to each sensor data;

[0160] an interpolation adjustment unit configured to adjust time series deviation by an interpolation method, and fill or delete data points exceeding a preset threshold;

[0161] a missing data repair unit configured to fill a missing part by using a mean value of front and rear data points.

[0162] As an implementation manner in the embodiment, the data preprocessing module comprises:

[0163] a smoothing filter unit configured to eliminate abnormal fluctuations in original data;

[0164] a wavelet decomposition unit configured to decompose the smoothed data into a high-frequency sub-signal and a low-frequency sub-signal;

[0165] a noise suppression unit configured to suppress noise components exceeding a preset threshold in the high-frequency sub-signal;

[0166] a signal reconstruction unit configured to integrate the suppressed sub-signals into a multi-dimensional data matrix after noise elimination.

[0167] As an implementation manner in the embodiment, the data standardization module comprises:

[0168] a dimension conversion unit configured to convert velocity, acceleration and displacement into a unified unit system;

[0169] a numerical mapping unit configured to map sensor data of different numerical ranges to a preset dimensionless interval through linear transformation;

[0170] a smoothing adjustment unit configured to correct the trend of data points deviating from the preset fluctuation range.

[0171] As an embodiment in the present embodiment, the data fusion module comprises:

[0172] a weight distribution unit configured to distribute initial weights according to the contribution degrees of acceleration, displacement and speed meter data;

[0173] a weight correction unit configured to dynamically adjust the weight proportion exceeding the preset threshold range;

[0174] a weighted fusion unit configured to fuse the standardized data set according to the adjusted weights to generate the reduced dimension feature vector.

[0175] As an embodiment in the present embodiment, the dynamic trend analysis module comprises:

[0176] a time window division unit configured to divide the current intensity level into multiple signal segments;

[0177] a pattern matching unit configured to extract the fluctuation characteristics of the signal segments and match them with the seismic wave patterns in the historical database;

[0178] a trend tracking unit configured to analyze the dynamic evolution law of the matched abnormal segments.

[0179] The system or device is used to realize the functions of the methods in the above embodiments, each module in the system or device corresponds to each step in the method, and has been described in the method and will not be described here.

[0180] Through the above embodiments, the problem of automatic recognition of seismic waves based on multi-source sensors in the related art is solved, so that the problems existing in the prior art can be solved.

[0181] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-source sensor-based automatic seismic wave recognition method, characterized in that, The method comprises the following steps: Synchronously collecting data through a speed meter, an accelerometer and a displacement meter to form an initial multidimensional data set; Performing time alignment processing on the initial multidimensional data set to generate a multi-source data group of uniform time series; Performing denoising processing on the multi-source data group of uniform time series to generate a multidimensional data matrix after denoising; Performing data standardization processing on the multidimensional data matrix to form a standardized data set; Performing dimension reduction processing on the standardized data set through principal component analysis to extract a main feature vector and obtain a fused feature data set; Classifying the feature data set through a support vector machine classification algorithm to determine whether there is a seismic wave signal; Performing short-time Fourier transform on the feature data set with the seismic wave signal to obtain frequency distribution characteristics; Determining the intensity level of the seismic wave signal according to the frequency distribution characteristics and a preset threshold range; Comparing and analyzing the intensity level with historical data to generate a dynamic change trend of the seismic wave; Determining a warning level according to the dynamic change trend and outputting an alarm signal.

2. The method of claim 1, wherein, The time alignment processing process comprises: adjusting time series deviation through an interpolation method according to the time stamp of each sensor; if the time stamp deviation exceeds a preset threshold, interpolating and filling in missing data points and deleting redundant data points; and repairing missing values of the filled or deleted data group by filling in the missing part with the mean value of the front and rear data points.

3. The method of claim 1, wherein, The denoising processing process comprises: performing smoothing filtering on the multi-source data group of uniform time series to eliminate abnormal fluctuations; decomposing the smoothed data into high-frequency sub-signals and low-frequency sub-signals through wavelet transform; suppressing components with amplitudes exceeding a preset threshold in the high-frequency sub-signals; and reconstructing the suppressed high-frequency sub-signals and the low-frequency sub-signals into a multidimensional data matrix after denoising.

4. The method of claim 1, wherein, The data standardization processing process comprises: converting the dimensions of speed, acceleration and displacement into a uniform unit system; mapping sensor data of different numerical ranges to a preset dimensionless interval through linear transformation; and performing smoothing adjustment on data points deviating from the preset fluctuation range based on the trend of adjacent data points.

5. The method of claim 1, wherein, The principal component analysis process comprises: assigning an initial weight according to the contribution degree of the accelerometer, displacement and speed meter data; dynamically adjusting the weight proportion to make the sum equal to 1 if the weight exceeds a preset threshold range; and generating a dimension-reduced feature vector by weighting and fusing the standardized data set according to the adjusted weight.

6. The method of claim 1, wherein, The comparison and analysis process comprises: dividing the current intensity level into multiple signal segments according to a time window; extracting the intensity fluctuation characteristics of each signal segment and performing similarity matching with the seismic wave mode in the historical database; identifying an abnormal fluctuation segment according to the matching result and analyzing the dynamic evolution law of the abnormal segment through a trend tracking tool.

7. A multi-source sensor based seismic wave automatic identification system, characterized in that, The system comprises: A multi-source sensor data collection module for synchronously collecting data of a speed meter, an accelerometer and a displacement meter to generate an initial multidimensional data set; A time series alignment module for performing time alignment processing on the initial multidimensional data set to generate a multi-source data group of uniform time series; The data preprocessing module is configured to perform denoising processing on the multi-source data set of the uniform time sequence to generate a denoised multi-dimensional data matrix. The data standardization module is configured to perform data standardization processing on the multi-dimensional data matrix to form a standardized data set. The data fusion module is configured to perform dimension reduction processing on the standardized data set by principal component analysis to extract a main feature vector and generate a fused feature data set. The pattern recognition module is configured to perform classification on the feature data set by using a support vector machine classification algorithm to determine whether there is a seismic wave signal. The time-frequency analysis module is configured to perform short-time Fourier transform on the feature data set with the seismic wave signal to obtain frequency distribution characteristics. The signal strength determination module is configured to determine the intensity level of the seismic wave signal according to the frequency distribution characteristics and a preset threshold range. The dynamic trend analysis module is configured to compare and analyze the intensity level with historical data to generate a dynamic change trend of the seismic wave. The early warning information generation module is configured to determine a warning level according to the dynamic change trend and output an alarm signal.

8. The system of claim 7, wherein, The time sequence alignment module includes: The timestamp processing unit is configured to attach a uniform time identifier to each sensor data. The interpolation adjustment unit is configured to adjust time sequence deviation by using an interpolation method, and fill or delete data points that exceed a preset threshold. The missing data repair unit is configured to fill missing parts by using the mean of the front and rear data points.

9. A computer terminal device, characterized by The computer program is executed by the processor to implement the method for automatically identifying a seismic wave based on a multi-source sensor according to any one of claims 1-6. The computer program is executed by the processor to implement the method for automatically identifying a seismic wave based on a multi-source sensor according to any one of claims 1-6. ​ ​ 10. A computer-readable storage medium having stored thereon a computer program, characterized in that, ​

Images