Blasting and micro-seismic signal automatic identification method and system based on multi-dimensional feature fusion

Abstract

The invention provides a blasting and micro-seismic signal automatic identification method and system based on multi-dimensional feature fusion, and relates to the technical field of micro-seismic monitoring, and the method comprises the steps: carrying out the event detection of original waveform data based on a long time window STA / LTA algorithm and an AIC algorithm, and obtaining a starting point and an ending point of an event; extracting event characteristic parameters of multiple dimensions, wherein the event characteristic parameters comprise duration, an energy abrupt change ratio, a high-amplitude sample number, multi-channel synchronism and a wake-to-noise energy ratio; and based on the extracted event characteristic parameters, performing classification through a decision model, and identifying a blasting event and a micro-seismic event. According to the method, the integrity of blasting event detection can be effectively ensured, the recognition precision and robustness can be greatly improved through multi-dimensional feature fusion, and in addition, automatic and intelligent comprehensive research and judgment can be achieved.

Description

Technical Field

[0001] This invention relates to the field of microseismic monitoring technology, specifically to an automatic identification method and system for blasting and microseismic signals based on multi-dimensional feature fusion. Background Technology

[0002] In microseismic monitoring of mines, tunnels, and other engineering projects, the short-time average to long-time average (STA / LTA) ratio algorithm is widely used for the automatic identification of vibration events. This algorithm typically uses a relatively short time window parameter (e.g., STA = 0.02 seconds, LTA = 0.08 seconds) to capture microseismic signals with weak energy and short duration.

[0003] However, the above methods exhibit significant limitations when engineering blasting occurs in the monitoring environment. Due to the high energy and long duration of blasting signals, fixed short-window parameters can lead to a complete blasting event being incorrectly segmented into multiple discrete microseismic events, resulting in fragmented event identification and numerous false alarms. This not only severely reduces data processing efficiency but also affects the accuracy of monitoring results, requiring substantial manual intervention for subsequent screening. Summary of the Invention

[0004] To address the technical problems in related technologies, this invention provides a method and system for automatic identification of blasting and microseismic signals based on multi-dimensional feature fusion.

[0005] To achieve the above objectives, the technical solution adopted by the present invention includes: According to a first aspect of the present invention, an automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion is provided, comprising the following steps: Step S1: Perform event detection on the original waveform data based on the long-window STA / LTA algorithm and AIC algorithm to obtain the start and end points of the events; Step S2: Extract event feature parameters from multiple dimensions, including duration, energy mutation ratio, number of high-amplitude samples, multi-channel synchronization, and wake-to-noise energy ratio; Step S3: Based on the extracted event feature parameters, classify them using a decision model to identify blasting events and microseismic events.

[0006] Optionally, in the long-window STA / LTA algorithm in step S1, the STA window is 0.07 to 0.09 seconds and the LTA window is 0.28 to 0.36 seconds; In step S1, the AIC algorithm is used to accurately locate the arrival of P-waves near the event trigger point detected by STA / LTA.

[0007] Optionally, in step S2, the duration is defined as the total duration from the start point determined by the AIC algorithm to the end point determined by the long-window STA / LTA algorithm.

[0008] Optionally, in step S2, the energy mutation ratio is defined as the ratio of the average absolute value amplitude after returning to zero within a window following the AIC starting point to the value within a window of the same length before the starting point.

[0009] Optionally, in step S2, the number of high-amplitude samples is defined as the number of sampling points whose absolute amplitude exceeds a preset threshold during the entire event duration.

[0010] Optionally, in step S2, the multi-channel synchronization is defined as the number of sensor channels that are synchronously triggered in the monitoring network within a set short time window.

[0011] Optionally, in step S2, the wake wave to noise energy ratio is defined as the ratio of the zeroed absolute mean amplitude within a window after the event end point determined by the STA / LTA algorithm to the zeroed absolute mean amplitude within a window of the same length before the precise start point of AIC.

[0012] According to a second aspect of the present invention, an automatic identification system for blasting and microseismic signals based on multi-dimensional feature fusion is also provided, for implementing the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any of the technical solutions of the first aspect of the present invention, comprising: The event detection module is used to perform event detection on the raw waveform data based on the long-window STA / LTA algorithm and the AIC algorithm to obtain the start and end points of the events; The feature extraction module is used to extract event feature parameters from multiple dimensions; The classification module is used to classify blasting events and microseismic events based on the extracted event feature parameters through a decision model.

[0013] According to a third aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, it is able to implement the steps of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any of the technical solutions of the first aspect of the present invention.

[0014] According to a fourth aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, is capable of implementing the steps of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any of the technical solutions of the first aspect of the present invention.

[0015] Beneficial effects: 1. Through the above technical solution, firstly, the method of the present invention can fundamentally solve the problem of event "fragmentation" misjudgment and improve the completeness of event detection. Specifically, the long-window STA / LTA algorithm and AIC algorithm adopted in the present invention are fundamentally different from the short-window algorithms used in existing related technologies for detecting microseismic events. This design is specifically designed for the physical characteristics of long-duration blasting signals. The longer window can smooth out brief noise fluctuations and respond to continuous energy release, thereby identifying the entire blasting process (including the main blast and subsequent vibrations) as a single, continuous event. In this way, from the perspective of algorithm principle, the fragmentation problem of traditional short-window methods incorrectly dividing a complete blasting into multiple microseismic events can be effectively overcome, laying a solid foundation for subsequent accurate analysis.

[0016] Second, the method of this invention can overcome the limitations of single feature criteria, effectively improving the accuracy and robustness of identification. Specifically, in step S2 of this invention, a set of feature parameters containing five specific dimensions is extracted, explicitly abandoning the criterion of relying on a single energy ratio threshold compared with existing related technologies. The five features selected in this invention (duration, energy mutation ratio, number of high-amplitude samples, multi-channel synchronization, and wake-to-noise energy ratio) comprehensively characterize the intrinsic differences between blasting and microseismic signals from different physical dimensions such as time domain, energy intensity, spatial distribution, and attenuation characteristics. A single feature may fail under certain boundary conditions, but multiple features constitute a multi-dimensional "chain of evidence," enabling the system to cope with more complex real-world scenarios.

[0017] For example, a long duration alone may not be sufficient for a determination, but if multiple characteristics are simultaneously met, such as a dramatic energy change (high energy change ratio) and simultaneous triggering by multiple sensors (strong multi-channel synchronization), the confidence level for it being an explosion event will be greatly increased. Thus, this multi-feature fusion strategy can effectively enhance the accuracy and robustness of classification.

[0018] Third, the method of this invention enables a leap from simple threshold judgment to intelligent comprehensive analysis, improving the automation and reliability of the method. Specifically, the method of this invention is based on extracted multi-dimensional features and classifies them through a decision model. This upgrades the method from the rigid "if-else" threshold judgment of existing technologies to a comprehensive decision system capable of handling complex feature relationships. Whether the model is based on rule-based logic trees or machine learning algorithms, it can perform weighted, combined, and comprehensive analysis of the five-dimensional feature information, thereby making a more accurate and reliable judgment than any single feature. This not only reduces the stringent requirements for setting thresholds for single features but also makes the entire identification process more intelligent, reduces the need for manual verification, and improves the automation level and reliability of the entire monitoring process.

[0019] Overall, the method of this invention achieves significant technological progress through the organic combination of a long-window STA / LTA+AIC event detection method, a five-dimensional feature set, and a comprehensive decision model. The method of this invention can not only effectively ensure the integrity of explosion event detection, but also greatly improve the recognition accuracy and robustness through multi-dimensional feature fusion. In addition, it can also realize automated and intelligent comprehensive judgment.

[0020] 2. Other beneficial effects or advantages of the present invention will be described in detail in the specific embodiments. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] in: Figure 1 This is a schematic flowchart of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion provided by an exemplary embodiment of the present invention; Figures 2 to 5 This is a real blasting vibration waveform diagram provided by an exemplary embodiment of the present invention, wherein, Figure 2 This is the waveform diagram of the explosion vibration of sensor 01. Figure 3 This is the waveform diagram of the blasting vibration of sensor 07. Figure 4 This is the waveform diagram of the blasting vibration of sensor 08. Figure 5 This is the waveform diagram of the blasting vibration of sensor 09; Figures 6 to 9This is a schematic diagram illustrating how the STA / LTA parameters using microseismic analysis cannot accurately identify blasting events, provided by an exemplary embodiment of the present invention. Figure 6 This is the waveform diagram for sensor 01 anomaly detection. Figure 7 This is the waveform diagram for sensor 07 anomaly detection. Figure 8 This is the waveform diagram for sensor 08 anomaly detection. Figure 9 This is the waveform diagram for sensor 09 anomaly identification. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0024] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0025] Furthermore, the terms "comprising" and "having," and any variations thereof, used in the description of this invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices. It should also be noted that in embodiments of this invention, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in embodiments of this invention should not be construed as preferred or advantageous over other embodiments or designs. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0026] To facilitate a clearer and more accurate understanding of the technical solutions of this invention by those skilled in the art, the existing related technologies and their technical problems will be described in more detail below.

[0027] In fields such as mining, tunnel engineering, water conservancy construction, geothermal resource development, and geological disaster monitoring, microseismic monitoring technology is widely used as a key means of assessing the stability of geological bodies, providing early warning of rock mass instability and monitoring engineering-induced seismic activity. This technology relies on a high-sensitivity sensor array to capture and analyze weak vibrations in the strata in real time.

[0028] However, in real-world monitoring environments, engineering blasting is a common source of interference. The vibration waveforms generated by blasting are characterized by high energy and long duration, making them easily confused with the waveforms of target microseismic events. Existing automated microseismic event identification algorithms, such as triggering algorithms based on the short-time to long-time average (STA / LTA) ratio, typically employ short analysis window parameters to capture weak, short-lived microseismic signals. This fixed, short parameter setting has significant limitations when dealing with blasting events. It may lead to a complete blasting event being incorrectly segmented and identified as multiple independent microseismic events, or generate a large number of false alarms. These problems severely impact the processing efficiency and accuracy of microseismic monitoring data, and may even lead to incorrect assessments of engineering safety conditions.

[0029] The limitations of traditional algorithms include: First, traditional microseismic detection algorithms (such as STA / LTA with standard parameters) often incorrectly identify blasting signals as multiple discrete, short-term microseismic events due to the long duration and complex energy distribution of the blasting waveform, leading to false alarms and event segmentation errors.

[0030] Second, the identification accuracy is low. Traditional microseismic detection algorithms lack comprehensive analysis of the intrinsic characteristics of the signal and rely solely on a single energy ratio threshold, which cannot effectively distinguish between blasting and microseismic signals that have similar triggering characteristics but are fundamentally different.

[0031] Third, the level of automation is low. False alarms and false negatives in traditional microseismic detection algorithms require extensive manual verification and screening, resulting in low efficiency and failing to meet the needs of real-time, automated monitoring.

[0032] In view of this, the present invention provides a novel solution: an automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion. The technical concept of this invention lies in abandoning the traditional approach of attempting to identify all types of events using a single, universal algorithm parameter, and instead adopting a targeted strategy tailored to the specific situation. Specifically, firstly, considering the strong energy and long duration of blasting signals, an event detection process combining a long-window STA / LTA algorithm and an AIC algorithm is designed. This aims to ensure that blasting events are captured completely and accurately as a whole from the source, thereby eradicating the persistent problem of "event fragmentation." Based on this, instead of relying on a single energy threshold, a set of multi-dimensional features strongly correlated with the blasting physical mechanism (such as duration, energy mutation, number of high-amplitude samples, multi-channel synchronization, etc.) is extracted from the complete waveform. Finally, a comprehensive judgment is made through a decision rule model based on clear physical meaning. This concept achieves a leap from "rough perception" to "precise profiling," fundamentally improving the accuracy, interpretability, and engineering practicality of identification.

[0033] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

[0034] like Figure 1 As shown, according to a first aspect of the present invention, an automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion is provided, comprising the following steps: Step S1: Perform event detection on the original waveform data based on the long-window STA / LTA algorithm and AIC algorithm to obtain the start and end points of the events; Step S2: Extract event feature parameters from multiple dimensions, including duration, energy mutation ratio, number of high-amplitude samples, multi-channel synchronization, and wake-to-noise energy ratio; Step S3: Based on the extracted event feature parameters, classify them using a decision model to identify blasting events and microseismic events.

[0035] Through the above technical solutions, firstly, the method of the present invention can fundamentally solve the problem of event "fragmentation" misjudgment and improve the completeness of event detection. Specifically, the long-window STA / LTA algorithm and AIC algorithm adopted in the present invention are fundamentally different from the short-window algorithms used in existing related technologies for detecting microseismic events. This design is specifically designed for the physical characteristics of long-duration blasting signals. The longer window can smooth out brief noise fluctuations and respond to continuous energy release, thereby identifying the entire blasting process (including the main blast and subsequent vibrations) as a single, continuous event. In this way, from the perspective of algorithm principle, the fragmentation problem of traditional short-window methods erroneously dividing a complete blasting into multiple microseismic events can be effectively overcome, laying a solid foundation for subsequent accurate analysis.

[0036] Second, the method of this invention can overcome the limitations of single feature criteria, effectively improving the accuracy and robustness of identification. Specifically, in step S2 of this invention, a set of feature parameters containing five specific dimensions is extracted, explicitly abandoning the criterion of relying on a single energy ratio threshold compared with existing related technologies. The five features selected in this invention (duration, energy mutation ratio, number of high-amplitude samples, multi-channel synchronization, and wake-to-noise energy ratio) comprehensively characterize the intrinsic differences between blasting and microseismic signals from different physical dimensions such as time domain, energy intensity, spatial distribution, and attenuation characteristics. A single feature may fail under certain boundary conditions, but multiple features constitute a multi-dimensional "chain of evidence," enabling the system to cope with more complex real-world scenarios.

[0037] For example, a long duration alone may not be sufficient for a determination, but if multiple characteristics are simultaneously met, such as a dramatic energy change (high energy change ratio) and simultaneous triggering by multiple sensors (strong multi-channel synchronization), the confidence level for it being an explosion event will be greatly increased. Thus, this multi-feature fusion strategy can effectively enhance the accuracy and robustness of classification.

[0038] Third, the method of this invention enables a leap from simple threshold judgment to intelligent comprehensive analysis, improving the automation and reliability of the method. Specifically, the method of this invention is based on extracted multi-dimensional features and classifies them through a decision model. This upgrades the method from the rigid "if-else" threshold judgment of existing technologies to a comprehensive decision system capable of handling complex feature relationships. Whether the model is based on rule-based logic trees or machine learning algorithms, it can perform weighted, combined, and comprehensive analysis of the five-dimensional feature information, thereby making a more accurate and reliable judgment than any single feature. This not only reduces the stringent requirements for setting thresholds for single features but also makes the entire identification process more intelligent, reduces the need for manual verification, and improves the automation level and reliability of the entire monitoring process.

[0039] Overall, the method of this invention achieves significant technological progress through the organic combination of a long-window STA / LTA+AIC event detection method, a five-dimensional feature set, and a comprehensive decision model. The method of this invention can not only effectively ensure the integrity of explosion event detection, but also greatly improve the recognition accuracy and robustness through multi-dimensional feature fusion. In addition, it can also realize automated and intelligent comprehensive judgment.

[0040] The method of the present invention will be further described below with reference to an exemplary embodiment.

[0041] Step 1: Robust event detection and integrity capture based on long time windows.

[0042] A long-time-window STA / LTA algorithm and an AIC algorithm specifically designed for identifying explosions are used to perform preliminary event detection on the raw waveform data.

[0043] For the long-window STA / LTA algorithm, a significantly increased LTA window is used (e.g., from 0.08 seconds to 0.32 seconds or longer). This makes the calculation of background noise more stable, effectively suppressing short-term fluctuations caused by pre-blast preparation activities or environmental noise, and preventing premature or erroneous triggering of the algorithm. Furthermore, the STA window is correspondingly increased (e.g., from 0.02 seconds to 0.08 seconds). This smooths out short-term energy peaks, ensuring that the algorithm only responds to persistent, significant energy changes, thus identifying the entire blasting event (including the main blast and subsequent secondary vibrations) as a single, continuous event, effectively solving the "event fragmentation" problem of traditional algorithms.

[0044] For the AIC algorithm, near the event trigger point detected by STA / LTA, the Akaike Information Criterion (AIC) algorithm is used for high-precision secondary localization of the P-wave arrival time (the true starting point of the event). The trigger point of the STA / LTA algorithm is usually located inside the waveform with significantly enhanced energy, rather than the true physical starting point. By calculating the goodness of fit of the signal autoregressive model, the AIC algorithm can accurately pinpoint the inflection point of the signal mode transition in the early stages of energy change. In this way, a precise time reference can be provided for subsequent feature calculations, avoiding feature distortion caused by trigger point delay.

[0045] The second step is the extraction and quantification of multi-dimensional feature parameters.

[0046] This step is crucial in distinguishing between blasting and microseismic events. After determining the precise start and end points of the event, the system automatically extracts the following set of key characteristic parameters that characterize the blasting properties.

[0047] For the first event feature parameter - duration (referred to as feature A), it is defined as: the total duration from the start point determined by AIC to the end point determined by the long time window STA / LTA.

[0048] Correspondingly, explosive events typically have a significantly long duration (e.g., >0.5 seconds) due to their long energy release process.

[0049] For the second event characteristic parameter - energy mutation ratio (referred to as characteristic B), it is defined as: the ratio of the average absolute value amplitude after returning to zero within a certain window (such as 300 points) after the AIC starting point to the value within the same length window before the starting point.

[0050] The corresponding explosive characteristics are: the energy of the explosive signal is instantaneous and enormous, with an extremely high signal-to-noise ratio. This characteristic value can increase by hundreds or even thousands of times.

[0051] For the third event characteristic parameter – the number of high-amplitude samples (referred to as feature C), it is defined as: during the entire event duration, the absolute value of the amplitude exceeds a preset high threshold (e.g., 0.299999, 0.3 m / s at full scale of the sensor). 2 The number of sampling points.

[0052] Corresponding blasting characteristics: blasting produces strong ground vibrations, causing sensors to record a large number of high-amplitude or saturated-amplitude sample points.

[0053] For the fourth event characteristic parameter - multi-channel synchronization (referred to as characteristic D), it is defined as: within a set extremely short time window (such as tens of milliseconds), how many sensor channels in the monitoring network simultaneously trigger an event that meets the above characteristics.

[0054] The corresponding explosive characteristics: An explosive event is a concentrated release of energy at or near the ground surface, and the resulting shock wave will reach all sensors in a region almost simultaneously. Therefore, an explosive event must be characterized by the synchronous triggering of multiple channels (or even all channels).

[0055] For the fifth event characteristic parameter - the wake-to-noise energy ratio (referred to as characteristic E), it is defined as: the "zero-return absolute mean amplitude" within a window (e.g., 300 sampling points) after the event end point determined by the STA / LTA algorithm, and the ratio of it to the "zero-return absolute mean amplitude" (i.e., background noise level) within a window of the same length before the precise start point of AIC.

[0056] Correspondingly, blasting events produce high-energy, long-lasting wake waves. Therefore, even after the STA / LTA indicates the event has ended, the actual vibration level remains significantly higher than the background noise level before the event. This ratio is much greater than 1 in blasting events, while it is closer to 1 in microseismic events.

[0057] The third step is intelligent classification based on decision rules.

[0058] Establish a multi-dimensional feature decision model to comprehensively evaluate the extracted features and finally output the classification result.

[0059] In this invention, the model can be a rule-based logical judgment (preferred) or a trained machine learning classifier (such as a decision tree, support vector machine, etc.).

[0060] For example, its rules can be: IF (Feature A: at least 2 channels have vibration durations > 0.5 seconds) AND (Feature B: At least 2 channels with energy mutation ratio > 500%) AND(Feature C: Number of channels exceeding the threshold amplitude >= 2) AND(Feature D: Number of synchronous trigger channels >= 3) AND (Feature E: Wake power to noise ratio of at least 2 channels >= 3.0) THEN Category result: "Explosion event" ELSE Classification result: "Microseismic event" or "Pending" In this way, the combination of multiple features and logic can greatly improve the accuracy and robustness of classification and avoid misjudgments that may be caused by a single feature.

[0061] The method of the present invention will be further described below with reference to a specific embodiment.

[0062] 1. Figures 2 to 5 These are waveforms (3000 points) from four channels in the actual blasting data of a certain project at 20:13:46 on August 8, 2025.

[0063] 2. Using the STA / LTA parameters of microseismic data cannot accurately identify blasting events.

[0064] Figures 6 to 9 The solid black vertical line in the diagram represents the start point of STA / LTA vibration, and the dashed black vertical line represents the end point. The STA / LTA parameters used (typical microseismic identification parameters) are as follows: Sampling rate: 2000Hz; STA window: 0.02 seconds (40 sampling points); LTA window: 0.08 seconds (160 sampling points); Trigger threshold: 2.5; Exit threshold: 0.5.

[0065] Under these parameters, the algorithm of this invention detected the following vibration events in the explosion waveform: sensor01: 2 vibration events, the first one is located at 1325-1760; the second one is located at 2218-2342. sensor07: 2 vibration events, the first one is located between 1353 and 1863; the second one is located between 2116 and 2369. sensor08: 3 vibration events (the first vibration at position 375-532 was discarded), the first position is 1353-1588; the second position is 1606-1786; the third position is 2221-2359; sensor09: 1 vibration event, the first one is located between 1351 and 1780.

[0066] It is clear that the STA / LTA parameters of micro-vibrations cannot accurately identify complete vibrations.

[0067] 3. Use blasting STA / LTA parameters to identify vibrations.

[0068] The program successfully identified the explosion event based on the explosion parameters (STA=0.08 seconds, LTA=0.32 seconds).

[0069] Parameter adjustment comparison: Original parameters (microseismic): STA = 0.02 seconds (40 points), LTA = 0.08 seconds (160 points); New parameters (detonation): STA = 0.08 seconds (160 points), LTA = 0.32 seconds (640 points); Vibration recognition results: sensor01: 1 vibration, starting position is 1325 (STA / LTA ratio: 2122.204), ending position is 2535 (STA / LTA ratio: 0.499), duration is 1210 sampling points (0.605 seconds), maximum STA / LTA ratio is 164564.43; sensor07: 1 vibration, starting position is 1353 (STA / LTA ratio: 27.409), ending position is 2655 (STA / LTA ratio: 0.500), duration is 1302 sampling points (0.651 seconds), maximum STA / LTA ratio is 10902.629; sensor08: 1 vibration, starting position is 1353 (STA / LTA ratio: 6.617), ending position is 2569 (STA / LTA ratio: 0.486), duration is 1216 sampling points (0.608 seconds), maximum STA / LTA ratio is 2676.446; sensor09: 1 vibration, starting position is 1352 (STA / LTA ratio: 840.579), ending position is 2478 (STA / LTA ratio: 0.498), duration is 1126 sampling points (0.563 seconds), maximum STA / LTA ratio is 83572.209; 4. Use the AIC algorithm to find the precise starting point.

[0070] AIC search range: All channels use a search window of 250 sampling points (200 points before the trigger point + 50 points after the trigger point); sensor01: AIC will refine the starting point from 1325 to 1324 (one sampling point earlier); sensor07: AIC confirms the starting point is 1353 (consistent with STA / LTA); sensor08: AIC confirms the starting point is 1353 (consistent with STA / LTA); sensor09: AIC will refine the starting point from 1352 to 1351 (one sampling point earlier); 5. The statistical data of the threshold amplitude ≥ 0.299999 m / s² for each channel are shown in Table 1 below.

[0071] The full-scale amplitude of the actual accelerometer is 0.3 m / s², so the high amplitude threshold is configured to be 0.299999 m / s².

[0072] Table 1. Number of over-threshold sampling points for each sensor 6. AIC Precise Start Point and Absolute Value Average Amplitude Analysis After Zeroing Table 2. AIC Precise Starting Point and Average Absolute Amplitude After Zeroing 7. Comprehensive judgment.

[0073] Feature A requires that the duration of all channels be greater than 0.5 seconds. Since the vibration durations of the four channels are [0.605 seconds, 0.651 seconds, 0.608 seconds, 0.563 seconds], they meet Feature A.

[0074] Feature B requires that at least two channels have an energy mutation ratio >500%, because the maximum STA / LTA ratio of the four channels is [164564.43, 10902.629, 2676.446, 83572.209], which meets the requirements of Feature B.

[0075] Feature C requires that the number of channels with over-threshold amplitude be greater than or equal to 2. Since the over-threshold amplitude array for four channels is [81,3,17,65], it meets the requirements of feature C.

[0076] Feature D requires that the number of synchronous triggering channels be greater than or equal to 3. Obviously, all 4 channels have STA / LTA triggering in the burst waveform, so they meet the requirements of feature D.

[0077] Feature E requires that the energy ratio of the wake wave to the noise in at least two channels be >= 3.0. The amplitude change ratio of the four sensors is [27.0949, 6.8999, 3.9083, 30.0303], which obviously meets the requirements of feature E.

[0078] Therefore, based on comprehensive analysis, this tremor was an explosion, not a micro-tremor.

[0079] In conclusion, using artificial intelligence (AI), particularly machine learning and deep learning models, to distinguish microseismic types is indeed a popular research area in both academia and industry. These methods are generally divided into two categories: One type of approach is based on "manual features + machine learning". Researchers first manually extract a series of features from the waveform (such as frequency domain features, time domain features, energy distribution, etc.), and then input these features into classifiers such as support vector machines (SVM) and random forests for training and classification.

[0080] Another type is the "end-to-end" deep learning method, which directly inputs the original waveform or its time-frequency spectrum into deep learning models such as convolutional neural networks (CNN) or recurrent neural networks (RNN), allowing the model to automatically learn and extract features and complete the classification, eliminating the need for manual feature design.

[0081] Although artificial intelligence methods can typically identify multiple types, compared to the method proposed in this invention, especially when considered in the context of practical engineering applications, this invention exhibits unique and significant advantages: First, interpretability. Classification methods in artificial intelligence generally suffer from the black-box problem, especially deep learning models. Their decision-making processes are extremely complex and difficult to explain using intuitive physical language. It's impossible to know precisely which specific shape or feature of the waveform the model bases its judgment on. This is a significant drawback in the field of safe production, where high reliability and post-event traceability are crucial. In contrast, the multi-dimensional physical feature fusion method of this invention ensures that each step and each feature has a clear and interpretable geophysical meaning. The decision-making logic (long duration, multi-point synchronization, rapid energy input, and slow decay) is completely transparent, conforming to the cognitive logic of human experts, resulting in reliable, credible, and traceable results.

[0082] Second, data dependence. The effectiveness of AI models heavily relies on a large amount of high-quality, labeled training data. In real-world projects, obtaining tens of thousands of accurately labeled blasting samples is extremely difficult and expensive. Poor data quality or imbalanced samples directly lead to a decline in model performance. This invention, however, is knowledge-driven rather than data-driven. Its core logic is built upon a deep understanding of the fundamental physical differences between blasting and microseismic activity. It does not require a large amount of data for training; only a small number of representative samples are needed to calibrate or verify the thresholds of physical characteristics, making its deployment extremely easy.

[0083] Third, generalization ability and adaptability. A model trained in one mining area may experience a sharp decline in performance (i.e., poor generalization ability) when directly applied to another mining area with different geological conditions, blasting techniques, and monitoring equipment. The model needs to be retrained or undergo complex transfer learning to adapt to the new environment. In particular, models trained on tunnel excavation data are almost infeasible for use in mining areas. This invention, however, relies on universal physical laws that remain largely unchanged across different mining areas (e.g., blasting always lasts longer and has a wider impact range than micro-seismic events). Therefore, this method has strong plug-and-play characteristics and is highly adaptable to new environments, requiring only fine-tuning of configurable thresholds.

[0084] Fourth, deployment costs and computing resources. Training AI models, especially deep learning models, requires expensive GPU computing resources and specialized algorithm engineers. Model deployment and subsequent maintenance are also relatively complex, requiring high computing power. However, in tunnel excavation sites or underground coal mines, it is difficult to find reliable high-performance hosts available. In contrast, the algorithms involved in this invention (STA / LTA, AIC, average value calculation) are all classic and efficient, with extremely low requirements for computing resources. They can be easily deployed on ordinary industrial control computers or even embedded devices, achieving truly low-cost, lightweight, and real-time operation.

[0085] Fifth, reliability and stability. AI models may become unpredictable when faced with novel signals or strong noise interference not present in the training set, sometimes making basic errors, and automatically generated reports may be unacceptable. This method, however, is a deterministic logical system. For a given input, its output is unique and deterministic. Its performance boundaries are clear, and the conditions under which it might fail can be predicted through analysis, ensuring stable and reliable system behavior.

[0086] According to a second aspect of the present invention, an automatic identification system for blasting and microseismic signals based on multi-dimensional feature fusion is also provided, for implementing the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any of the technical solutions of the first aspect of the present invention, comprising an event detection module, a feature extraction module, and a classification module. The event detection module is used to perform event detection on the original waveform data based on the long-window STA / LTA algorithm and the AIC algorithm to obtain the start and end points of the events; the feature extraction module is used to extract event feature parameters in multiple dimensions; and the classification module is used to classify and identify blasting events and microseismic events based on the extracted event feature parameters using a decision model.

[0087] Through this technical solution, the system of this invention achieves significant technological advancements by organically combining a long-window STA / LTA+AIC event detection method, a five-dimensional feature set, and a comprehensive decision model. The method not only effectively ensures the integrity of blasting event detection but also greatly improves recognition accuracy and robustness through multi-dimensional feature fusion. Furthermore, it enables automated and intelligent comprehensive judgment. The system of this invention solidifies the efficient and accurate method of this invention into a stable and reliable physical or software entity, enabling automated and integrated processing, effectively improving the efficiency and convenience of engineering applications, and ensuring the consistency and repeatability of recognition results.

[0088] According to a third aspect of the present invention, a computer device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it is able to implement the steps of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion in any of the technical solutions of the first aspect of the present invention.

[0089] It is understood that in this embodiment, the memory may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; furthermore, the memory may include combinations of the above types of memory. The present invention does not specifically limit this.

[0090] Similarly, the processor can implement or execute the various exemplary logical steps described in conjunction with the disclosure of this invention. The processor can be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logical steps described in conjunction with the disclosure of this invention. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0091] According to a fourth aspect of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, is capable of implementing the steps of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion in any of the technical solutions of the first aspect of the present invention.

[0092] In this embodiment, the computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof, or any other form of computer-readable storage medium known in the art. An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium may also be a component of the processor. The processor and the storage medium may reside in an application-specific integrated circuit (ASIC). In embodiments of the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0093] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion, characterized in that, Includes the following steps: Step S1: Perform event detection on the original waveform data based on the long-window STA / LTA algorithm and AIC algorithm to obtain the start and end points of the events; Step S2: Extract event feature parameters from multiple dimensions, including duration, energy mutation ratio, number of high-amplitude samples, multi-channel synchronization, and wake-to-noise energy ratio; Step S3: Based on the extracted event feature parameters, classify them using a decision model to identify blasting events and microseismic events.

2. The automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion according to claim 1, characterized in that, In the long-window STA / LTA algorithm in step S1, the STA window is 0.07 to 0.09 seconds, and the LTA window is 0.28 to 0.36 seconds. In step S1, the AIC algorithm is used to accurately locate the arrival of P-waves near the event trigger point detected by STA / LTA.

3. The automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion according to claim 1, characterized in that, In step S2, the duration is defined as the total duration from the start point determined by the AIC algorithm to the end point determined by the long-window STA / LTA algorithm.

4. The automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion according to claim 1, characterized in that, In step S2, the energy mutation ratio is defined as the ratio of the average absolute value amplitude after returning to zero within a window following the AIC starting point to the value within a window of the same length before the starting point.

5. The automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion according to claim 1, characterized in that, In step S2, the number of high-amplitude samples is defined as the number of sampling points whose absolute amplitude exceeds a preset threshold during the entire event duration.

6. The automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion according to claim 1, characterized in that, In step S2, the multi-channel synchronization is defined as the number of sensor channels that are synchronously triggered in the monitoring network within a set short time window.

7. The automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion according to claim 1, characterized in that, In step S2, the wake wave to noise energy ratio is defined as the ratio of the zeroed absolute mean amplitude within a window after the event end point determined by the STA / LTA algorithm to the zeroed absolute mean amplitude within a window of the same length before the precise start point of AIC.

8. An automatic identification system for blasting and microseismic signals based on multi-dimensional feature fusion, used to implement the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any one of claims 1-7, characterized in that, include: The event detection module is used to perform event detection on the raw waveform data based on the long-window STA / LTA algorithm and the AIC algorithm to obtain the start and end points of the events; The feature extraction module is used to extract event feature parameters from multiple dimensions; The classification module is used to classify blasting events and microseismic events based on the extracted event feature parameters through a decision model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it can implement the steps of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it can implement the steps of the automatic identification method for blasting and microseismic signals based on multi-dimensional feature fusion as described in any one of claims 1-8.

Images