Four-sensor-based regional major engineering earthquake precise alarm system and method

Abstract

The application discloses a four-sensor-based regional major engineering earthquake precise alarm system and method, which comprises the following steps: first, four acceleration sensors are arranged, original acceleration signals are synchronously collected and preprocessed to obtain four processed signals; second, local characteristic values of the processed signals are extracted and compared with threshold values, and a pre-alarm time is recorded when the local characteristic values all meet the threshold values; third, waveform similarity between the processed signal pairs is calculated in groups, and judgment is made according to the spatial correlation of seismic waves; fourth, the pre-alarm time difference between the processed signal pairs is calculated, and it is judged whether the time difference is within the time difference range of the seismic waves; if the time difference is within the time difference range of the seismic waves, the earthquake event is determined, and the earthquake pre-alarm is given; and fifth, after the earthquake pre-alarm, the processed signals are continuously monitored, and the earthquake alarm is given when all the processed signals exceed the earthquake alarm threshold. Therefore, the false alarm rate of the regional major engineering earthquake caused by regional interference is low.

Description

Technical Field

[0001] This invention relates to an earthquake alarm system and method, belonging to the field of earthquake monitoring and disaster early warning technology for major engineering projects, and particularly to a precise earthquake alarm system and method for regional major engineering projects based on four sensors. Background Technology

[0002] Major regional projects (such as nuclear power plants, large power transmission and transformation hubs, reservoirs and dams, key nodes of high-speed railways, and strategic energy facilities) are core components of the national economic lifeline and public safety. These projects are complex in structure and involve huge investments; once they suffer a destructive earthquake, they are prone to triggering catastrophic secondary disasters. Therefore, reliable earthquake early warning for major regional projects is crucial. False alarms in earthquake early warning systems for major regional projects can lead to unnecessary emergency responses, production interruptions, and public panic, causing economic losses and undermining public trust. Missed alarms mean missing valuable time to initiate critical safety measures such as emergency shutdown and high-speed braking, potentially resulting in facility damage and casualties.

[0003] Currently, P-wave identification methods based on vibration amplitude thresholds or signal feature detection (such as the STA / LTA algorithm) are commonly used in engineering sites to achieve early earthquake warnings, and this is a prerequisite for earthquake alarms. However, major engineering sites and their surrounding environments often contain a large number of strong non-seismic vibration sources (such as heavy vehicles, large machinery, construction blasting, etc.). The vibration signals generated by these disturbances are highly similar to the P-waves in the early stages of an earthquake in terms of amplitude and spectral characteristics, making it difficult for traditional methods to effectively distinguish between real earthquakes and local disturbances. This results in a high false alarm rate for major regional engineering earthquakes caused by regional disturbances.

[0004] Chinese patent application No. 201710071925.9, filed on April 26, 2017, discloses a method for earthquake early warning on high-speed railways, earthquake monitoring stations, and a railway bureau central system. The method involves collecting seismic waveform information from multiple seismic directional measurements at various seismic monitoring stations, performing similarity checks on the seismic waveform information, and if the check result exceeds a false alarm threshold, plotting a correlation curve of the seismic waveform information and determining the first arrival time of the P-wave based on the correlation curve. Each seismic monitoring station generates P-wave early warning information based on the seismic waveform information from multiple seismic directional measurements and sends the P-wave early warning information to the railway bureau central system, allowing the railway bureau central system to determine whether to issue an early warning and take appropriate action based on the P-wave early warning information. While this patent can improve the accuracy and timeliness of P-wave early warnings from stations along high-speed railways, it still has the following drawbacks:

[0005] This design avoids false alarms by triggering multiple stations together. Since the distance between each station is about 25km, the interference vibration is unlikely to affect multiple stations. The railway bureau's central system can determine whether it is an earthquake by the early warning results of multiple stations. However, the monitoring range of major regional projects is relatively concentrated. Multiple stations are easily triggered by the same local strong interference source at the same time, resulting in a high false alarm rate for earthquakes in major regional projects caused by regional interference. Furthermore, if a sensor at the same station fails, the similarity algorithm fails, and the entire high-speed rail station loses its early warning function.

[0006] The information disclosed in this background section is intended only to enhance the understanding of the overall background of this patent application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this invention is to overcome the defects and problems of high false alarm rates of regional major engineering earthquakes caused by regional interference in the existing technology, and to provide a precise alarm system and method for regional major engineering earthquakes based on four sensors with low false alarm rates caused by regional interference.

[0008] To achieve the above objectives, the technical solution of the present invention is: a regional major engineering earthquake precision alarm system and method based on four sensors, the system comprising four acceleration sensors, a first strong earthquake recorder, a second strong earthquake recorder, a central processing unit and an alarm box;

[0009] The four acceleration sensors are a first acceleration sensor, a second acceleration sensor, a third acceleration sensor, and a fourth acceleration sensor;

[0010] The first acceleration sensor and the second acceleration sensor are connected to the first strong earthquake recorder, and the third acceleration sensor and the fourth acceleration sensor are connected to the second strong earthquake recorder;

[0011] The first strong-motion recorder and the second strong-motion recorder are respectively connected to the alarm box via relays;

[0012] Meanwhile, both the first and second strong earthquake recorders are connected to the central processing unit, which in turn is connected to the alarm box.

[0013] The precise alarm method includes the following steps:

[0014] Step 1: Within the area covered by the major project, four accelerometers are evenly deployed; the four accelerometers collect data synchronously to obtain four raw acceleration signals, and each raw acceleration signal is preprocessed to obtain four processed signals, namely the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal.

[0015] Step 2: First, extract the local feature values ​​of the four processed signals respectively, and then compare the local feature values ​​of the four processed signals with the corresponding earthquake feature thresholds respectively; if the local feature values ​​of the four processed signals all meet the corresponding earthquake feature thresholds, then record an early warning trigger time for each processed signal and proceed to Step 3; otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0016] Step 3: First, perform waveform similarity calculations between the first and second processed signals, and between the third and fourth processed signals, to obtain the first similarity value and the second similarity value. Then, compare the first and second similarity values ​​with the seismic wave similarity judgment threshold. If both the first and second similarity values ​​meet the seismic wave similarity judgment threshold, proceed to step 4. Otherwise, it is determined that no earthquake event has occurred, and the early warning process terminates.

[0017] Step 4: First, based on the warning trigger time of each processed signal recorded in Step 2, calculate the first warning trigger time difference between the first and second processed signals, and the second warning trigger time difference between the third and fourth processed signals; then compare the first and second warning trigger time differences with the time difference range of seismic waves; if either the first or second warning trigger time difference does not conform to the time difference range of seismic waves, it is determined that no earthquake event has occurred, and the warning process terminates; if both the first and second warning trigger time differences conform to the time difference range of seismic waves, it is determined that an earthquake event has occurred, earthquake warning parameters are calculated, and an earthquake warning is issued.

[0018] Step 5: After issuing an earthquake early warning, continue to monitor the processed signals of the above-mentioned acceleration sensors; if the processed signal of any acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm will be issued and the alarm process will end; if the processed signals of all four acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm will be issued.

[0019] The precise alarm method also includes a fault-tolerant mode:

[0020] The system continuously monitors the signals from the four aforementioned accelerometer sensors. If any one of the accelerometer signals fails, it automatically switches to a three-sensor operating mode based on the remaining three valid sensor signals, while maintaining the first step described above. The three-sensor operating mode includes:

[0021] In the second step, the local feature values ​​of the three processed signals are extracted first, and then the local feature values ​​of the three processed signals are compared with the corresponding earthquake feature thresholds. If the local feature values ​​of the three processed signals all meet the corresponding earthquake feature thresholds, an early warning trigger time is recorded for each processed signal, and the process proceeds to the third step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0022] In the third step, the waveform similarity between each pair of the three processed signals is calculated to obtain the waveform similarity value between each pair. Then, the waveform similarity value between each pair is compared with the seismic wave similarity judgment threshold. If all the similarity values ​​between each pair meet the seismic wave similarity judgment threshold, the process proceeds to the fourth step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0023] In the fourth step, based on the warning trigger time of each processed signal recorded in the second step, the warning trigger time difference between each pair of the three processed signals is calculated. Then, all the warning trigger time differences between each pair are compared with the time difference range of the seismic waves. If any one of the warning trigger time differences between each pair does not conform to the time difference range of the seismic waves, it is determined that no earthquake event has occurred, and the warning process is terminated. If all the warning trigger time differences between each pair conform to the time difference range of the seismic waves, it is determined that an earthquake event has occurred, and earthquake warning parameters are calculated and an earthquake warning is issued.

[0024] In the fifth step, after issuing an earthquake early warning, the processed signals of the three acceleration sensors continue to be monitored; if the processed signal of any one acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm is issued and the alarm process ends; if the processed signals of all three acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm is issued.

[0025] In the first step, the four acceleration sensors are deployed in the free field. Each acceleration sensor is installed via an instrument pier, which is connected to the bedrock or old soil layer below the free field.

[0026] In the first step, the preprocessing includes noise reduction processing;

[0027] The noise reduction process involves performing a 0.1Hz-33Hz bandpass filter on each original acceleration signal.

[0028] In the second step, extracting the local feature values ​​of the four processed signals respectively, and then comparing the local feature values ​​of the four processed signals with the corresponding seismic feature thresholds, means:

[0029] Perform the following operations on the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal, respectively:

[0030] First, the STA / LTA algorithm is used to coarsely pick up the first arrival of the P-wave in the processed signal to obtain the coarse arrival time; then, the VAR-AIC algorithm is used to accurately pick up the first arrival of the P-wave within a preset time window, using the coarse arrival time as a reference point, to determine the first arrival time of the P-wave.

[0031] Starting from the initial arrival time of the P wave, waveform data for 1 second is extracted. Then, based on this waveform data, the number of zero crossings, kurtosis growth factor, and spectral energy ratio of the processed signal are calculated.

[0032] Finally, the number of zero-crossings of each processed signal is compared with the seismic feature threshold corresponding to the number of zero-crossings, the kurtosis growth coefficient of each processed signal is compared with the seismic feature threshold corresponding to the kurtosis growth coefficient, and the spectral energy ratio of each processed signal is compared with the seismic feature threshold corresponding to the spectral energy ratio.

[0033] The seismic feature threshold corresponding to the number of zero crossings is a preset upper frequency value; when the number of zero crossings of the processed signal is not greater than the upper frequency value, the processed signal meets the seismic feature threshold corresponding to the number of zero crossings.

[0034] The seismic characteristic threshold corresponding to the kurtosis growth coefficient is a preset upper limit of the rate of change; when the kurtosis growth coefficient of the processed signal is not greater than the upper limit of the rate of change, the processed signal meets the seismic characteristic threshold corresponding to the kurtosis growth coefficient.

[0035] The seismic characteristic threshold corresponding to the spectral energy ratio is a preset lower limit value; when the spectral energy ratio of the processed signal is not less than the lower limit value, the processed signal meets the seismic characteristic threshold corresponding to the spectral energy ratio.

[0036] In the fourth step, the time difference range of the seismic wave is the theoretical wave arrival time difference range calculated based on the preset P-wave propagation velocity, S-wave propagation velocity and the actual distance between each acceleration sensor.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] 1. This invention discloses a precise earthquake early warning system and method for major regional engineering projects based on four sensors. The method includes: firstly, deploying four spatially separated accelerometers around the engineering area and synchronously acquiring signals; obtaining first to fourth processed signals after preprocessing; then extracting local feature values ​​of each processed signal and determining whether they meet the corresponding earthquake feature thresholds; if the local feature values ​​of all four processed signals meet the thresholds, then recording the early warning trigger time for each signal; then, calculating the waveform similarity between the first and second processed signals, and between the third and fourth processed signals, respectively, to obtain first and second similarity values; and then comparing the first and second similarity values ​​with the seismic wave similarity. The judgment threshold comparison is performed. If both the first and second similarity values ​​are met, the first warning trigger time difference between the first and second processed signals and the second warning trigger time difference between the third and fourth processed signals are calculated based on the warning trigger time. Then, it is determined whether the first and second warning trigger time differences are within the time difference range of the seismic waves. If both the first and second warning trigger time differences are within the time difference range of the seismic waves, it is determined to be an earthquake event. Earthquake warning parameters are calculated and an earthquake warning is issued. After the earthquake warning, the processed signals of each acceleration sensor are monitored. When all processed signals exceed the preset earthquake alarm threshold, an earthquake alarm is issued. The advantages of this invention also include:

[0039] Firstly, by deploying four accelerometers in the engineering area, the distance between the sensors is increased, which reduces the probability that strong local interference (such as vehicles, machinery, etc.) will produce similar waveforms on multiple sensors, thereby improving the spatial identifiability and acquisition reliability of the signal source at the physical level.

[0040] Secondly, by performing local feature judgment on each processed signal, high-frequency mechanical vibrations and instantaneous impact interference that are similar to seismic waves only in a single dimension such as frequency or rate of change are effectively filtered out.

[0041] Thirdly, by calculating and judging the waveform similarity between multi-sensor signals, and based on the physical nature of the high spatial correlation of waveforms in real earthquake events, it effectively distinguishes between spatially consistent seismic waves and spatially random local concurrent interference, thereby improving the credibility of event identification.

[0042] Fourthly: By verifying whether the time difference of the early warning trigger of each sensor conforms to the time difference range of seismic wave propagation, and by utilizing the order-of-magnitude difference in the propagation speed between seismic waves and surface interference waves, the essential distinction of strong near-field interference is realized.

[0043] Fifthly, by judging local characteristics, waveform similarity, and time difference rationality, the ability to resist false alarms in complex interference environments is enhanced; thus ensuring that every warning corresponds to a real earthquake event and reducing the false alarm rate.

[0044] Therefore, this invention can not only provide earthquake early warning for major regional projects, but also has a low false alarm rate for earthquakes in major regional projects caused by regional interference.

[0045] 2. In the present invention, a regional major engineering earthquake precision alarm system and method based on four sensors, the method further includes a fault-tolerant mode. The system detects the signals of the four aforementioned acceleration sensors in real time. When any acceleration sensor signal fails, it automatically switches to a three-sensor operating mode based on the remaining three valid sensor signals, while maintaining the first step. In application, if any acceleration sensor fails (e.g., due to data interruption or abnormality), while maintaining the original deployment and acquisition architecture, all judgment conditions in steps two through five are adaptively and seamlessly switched from "four-sensor" logic to "three-sensor" logic. Specifically, in the second step, all local feature values ​​of the three valid processed signals must meet a threshold. In the third step, the waveform similarity between any two pairs of the three valid processed signals is calculated and determined. In the fourth step, the time difference between any two pairs of the three early warning trigger times is calculated and determined. After an earthquake early warning, the processed signals of the three valid acceleration sensors are monitored. When the processed signals of all valid sensors exceed the earthquake alarm threshold, an earthquake alarm is triggered. The advantages of this invention also include:

[0046] Firstly, when a single-point sensor failure occurs, the system seamlessly downgrades to a three-sensor mode to continue operating, ensuring that the core early warning and judgment capabilities remain uninterrupted. This completely eliminates the risk of missed alarms caused by the complete loss of early warning functions due to equipment failure, reduces the missed alarm rate, and improves reliability.

[0047] Secondly, in the three-sensor working mode, the early warning is not a simplified process, but rather the execution of a complete and strictly adapted set of logic, including the matching of three signal characteristics, the similarity of two pairs of waveforms, and the reasonable arrival time difference between two pairs. This ensures that even in extreme fault conditions, while avoiding missed alarms, high recognition accuracy can still be maintained, effectively preventing the increase in false alarm rate that may be caused by loose logic.

[0048] Therefore, this invention not only has a low false alarm rate for earthquakes in major regional engineering projects, but also reduces the false alarm rate.

[0049] 3. In the present invention, a precise earthquake alarm system and method for major regional engineering projects based on four sensors, in the second step, the local feature values ​​include the signal zero-crossing count, kurtosis growth coefficient, and spectral energy ratio. In application, the processed signals from each sensor are extracted and analyzed in parallel, including the three local feature values: signal zero-crossing count, kurtosis growth coefficient, and spectral energy ratio. The signal zero-crossing count is used to suppress high-frequency environmental vibrations, the kurtosis growth coefficient is used to identify transient impact interference, and the spectral energy ratio is used to distinguish between seismic waves dominated by low-frequency energy and mechanical vibrations dominated by high-frequency energy. By jointly judging and cross-validating these three local feature values, an essential leap from relying on a single amplitude to multi-dimensional physical characteristics is achieved. This allows for the precise filtering out of complex engineering interferences that are similar in amplitude but do not match seismic waves in essential characteristics such as frequency, degree of abrupt change, and energy distribution, thereby improving the anti-interference capability and reliability of single-point signal identification. Therefore, the present invention not only reduces the false alarm rate but also improves the anti-interference capability and reliability of single-point signal identification.

[0050] 4. In the present invention, a precise earthquake alarm system and method for major regional engineering projects based on four sensors, a first and a second accelerometer are connected to a first strong-motion recorder, and a third and a fourth accelerometer are connected to a second strong-motion recorder. The first and second strong-motion recorders are respectively connected to an alarm box via relays. In application, when all four accelerometers are functioning normally, the first and second strong-motion recorders synchronously complete signal preprocessing, local feature extraction, intra-group waveform similarity calculation, and trigger time difference judgment for the connected sensors. When both the first and second strong-motion recorders independently determine an earthquake event, they directly send a hardware trigger signal to the alarm box via relays. This pure hardware triggering mechanism bypasses software and communication links, eliminating processing delays and uncertainties at the source. Furthermore, due to its simple structure, it ensures stable and reliable early warning judgment while improving the system's early warning timeliness. Therefore, the present invention not only improves the anti-interference capability and reliability of single-point signal identification but also enhances the early warning timeliness.

[0051] 5. In the present invention, a precise earthquake alarm system and method for major regional engineering projects based on four sensors, the core discrimination logic of the method is based on multi-sensor waveform similarity analysis. Its essence lies in distinguishing between seismic events and non-seismic vibration events. This discrimination logic is independent of any specific engineering structure or operating mode. The sensor deployment scheme and signal processing algorithm of the method are universally applicable and suitable for various types of major regional engineering projects such as nuclear power plants, large-scale power transmission and transformation projects, and reservoir dams, thus possessing broad technical applicability and engineering promotion value. Therefore, the present invention not only improves the timeliness of early warning but also has universality and adaptability. Attached Figure Description

[0052] Figure 1 This is a system structure diagram of the present invention.

[0053] Figure 2 This is the earthquake early warning logic diagram under the four-sensor working mode in this invention.

[0054] Figure 3 This is a logic diagram of earthquake early warning under the three-sensor working mode in this invention.

[0055] In the diagram: First accelerometer 1, Second accelerometer 2, Third accelerometer 3, Fourth accelerometer 4, First strong earthquake recorder 5, Second strong earthquake recorder 6, Central processing unit 7, Alarm box 8. Detailed Implementation

[0056] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0057] See Figures 1-3 A precise earthquake alarm system and method for major regional engineering projects based on four sensors, the system comprising four acceleration sensors, a first strong earthquake recorder 5, a second strong earthquake recorder 6, a central processing unit 7, and an alarm box 8;

[0058] The four acceleration sensors are the first acceleration sensor 1, the second acceleration sensor 2, the third acceleration sensor 3, and the fourth acceleration sensor 4.

[0059] The first acceleration sensor 1 and the second acceleration sensor 2 are connected to the first strong earthquake recorder 5, and the third acceleration sensor 3 and the fourth acceleration sensor 4 are connected to the second strong earthquake recorder 6;

[0060] The first strong earthquake recorder 5 and the second strong earthquake recorder 6 are respectively connected to the alarm box 8 via relays;

[0061] Meanwhile, both the first strong earthquake recorder 5 and the second strong earthquake recorder 6 are connected to the central processing unit 7, which is also connected to the alarm box 8.

[0062] The precise alarm method includes the following steps:

[0063] Step 1: Within the area covered by the major project, four accelerometers are evenly deployed; the four accelerometers collect data synchronously to obtain four raw acceleration signals, and each raw acceleration signal is preprocessed to obtain four processed signals, namely the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal.

[0064] Step 2: First, extract the local feature values ​​of the four processed signals respectively, and then compare the local feature values ​​of the four processed signals with the corresponding earthquake feature thresholds respectively; if the local feature values ​​of the four processed signals all meet the corresponding earthquake feature thresholds, then record an early warning trigger time for each processed signal and proceed to Step 3; otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0065] Step 3: First, perform waveform similarity calculations between the first and second processed signals, and between the third and fourth processed signals, to obtain the first similarity value and the second similarity value. Then, compare the first and second similarity values ​​with the seismic wave similarity judgment threshold. If both the first and second similarity values ​​meet the seismic wave similarity judgment threshold, proceed to step 4. Otherwise, it is determined that no earthquake event has occurred, and the early warning process terminates.

[0066] Step 4: First, based on the warning trigger time of each processed signal recorded in Step 2, calculate the first warning trigger time difference between the first and second processed signals, and the second warning trigger time difference between the third and fourth processed signals; then compare the first and second warning trigger time differences with the time difference range of seismic waves; if either the first or second warning trigger time difference does not conform to the time difference range of seismic waves, it is determined that no earthquake event has occurred, and the warning process terminates; if both the first and second warning trigger time differences conform to the time difference range of seismic waves, it is determined that an earthquake event has occurred, earthquake warning parameters are calculated, and an earthquake warning is issued.

[0067] Step 5: After issuing an earthquake early warning, continue to monitor the processed signals of the above-mentioned acceleration sensors; if the processed signal of any acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm will be issued and the alarm process will end; if the processed signals of all four acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm will be issued.

[0068] The precise alarm method also includes a fault-tolerant mode:

[0069] The system continuously monitors the signals from the four aforementioned accelerometer sensors. If any one of the accelerometer signals fails, it automatically switches to a three-sensor operating mode based on the remaining three valid sensor signals, while maintaining the first step described above. The three-sensor operating mode includes:

[0070] In the second step, the local feature values ​​of the three processed signals are extracted first, and then the local feature values ​​of the three processed signals are compared with the corresponding earthquake feature thresholds. If the local feature values ​​of the three processed signals all meet the corresponding earthquake feature thresholds, an early warning trigger time is recorded for each processed signal, and the process proceeds to the third step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0071] In the third step, the waveform similarity between each pair of the three processed signals is calculated to obtain the waveform similarity value between each pair. Then, the waveform similarity value between each pair is compared with the seismic wave similarity judgment threshold. If all the similarity values ​​between each pair meet the seismic wave similarity judgment threshold, the process proceeds to the fourth step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0072] In the fourth step, based on the warning trigger time of each processed signal recorded in the second step, the warning trigger time difference between each pair of the three processed signals is calculated. Then, all the warning trigger time differences between each pair are compared with the time difference range of the seismic waves. If any one of the warning trigger time differences between each pair does not conform to the time difference range of the seismic waves, it is determined that no earthquake event has occurred, and the warning process is terminated. If all the warning trigger time differences between each pair conform to the time difference range of the seismic waves, it is determined that an earthquake event has occurred, and earthquake warning parameters are calculated and an earthquake warning is issued.

[0073] In the fifth step, after issuing an earthquake early warning, the processed signals of the three acceleration sensors continue to be monitored; if the processed signal of any one acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm is issued and the alarm process ends; if the processed signals of all three acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm is issued.

[0074] In the first step, the four acceleration sensors are deployed in the free field. Each acceleration sensor is installed via an instrument pier, which is connected to the bedrock or old soil layer below the free field.

[0075] In the first step, the preprocessing includes noise reduction processing;

[0076] The noise reduction process involves performing a 0.1Hz-33Hz bandpass filter on each original acceleration signal.

[0077] In the second step, extracting the local feature values ​​of the four processed signals respectively, and then comparing the local feature values ​​of the four processed signals with the corresponding seismic feature thresholds, means:

[0078] Perform the following operations on the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal, respectively:

[0079] First, the STA / LTA algorithm is used to coarsely pick up the first arrival of the P-wave in the processed signal to obtain the coarse arrival time; then, the VAR-AIC algorithm is used to accurately pick up the first arrival of the P-wave within a preset time window, using the coarse arrival time as a reference point, to determine the first arrival time of the P-wave.

[0080] Starting from the initial arrival time of the P wave, waveform data for 1 second is extracted. Then, based on this waveform data, the number of zero crossings, kurtosis growth factor, and spectral energy ratio of the processed signal are calculated.

[0081] Finally, the number of zero-crossings of each processed signal is compared with the seismic feature threshold corresponding to the number of zero-crossings, the kurtosis growth coefficient of each processed signal is compared with the seismic feature threshold corresponding to the kurtosis growth coefficient, and the spectral energy ratio of each processed signal is compared with the seismic feature threshold corresponding to the spectral energy ratio.

[0082] The seismic feature threshold corresponding to the number of zero crossings is a preset upper frequency value; when the number of zero crossings of the processed signal is not greater than the upper frequency value, the processed signal meets the seismic feature threshold corresponding to the number of zero crossings.

[0083] The seismic characteristic threshold corresponding to the kurtosis growth coefficient is a preset upper limit of the rate of change; when the kurtosis growth coefficient of the processed signal is not greater than the upper limit of the rate of change, the processed signal meets the seismic characteristic threshold corresponding to the kurtosis growth coefficient.

[0084] The seismic characteristic threshold corresponding to the spectral energy ratio is a preset lower limit value; when the spectral energy ratio of the processed signal is not less than the lower limit value, the processed signal meets the seismic characteristic threshold corresponding to the spectral energy ratio.

[0085] In the fourth step, the time difference range of the seismic wave is the theoretical wave arrival time difference range calculated based on the preset P-wave propagation velocity, S-wave propagation velocity and the actual distance between each acceleration sensor.

[0086] The following are supplementary descriptions of the present invention:

[0087] In the fifth step of this invention, the earthquake alarm threshold refers to the critical acceleration value that triggers an earthquake alarm, which is set comprehensively based on the seismic fortification standards of major regional projects, the safety limits of key structures and equipment, and historical earthquake and measured data of the project site.

[0088] The reason for choosing four accelerometer sensors in this invention is as follows: First, from the perspective of the hardware implementation necessity of the system architecture, the core warning and discrimination logic of this system relies on the mechanism of "dual-path independent judgment and hardware AND gate triggering," which requires that the sensors must be connected in pairs to their respective data acquisition units (strong earthquake recorders); each acquisition unit needs to connect at least two accelerometer sensors to complete the waveform similarity analysis within the group. Therefore, the system requires at least four accelerometer sensors to form two independent and fully functional judgment units. In contrast, the two-sensor scheme can only form one unit and cannot achieve a dual-path redundancy architecture; while the three-sensor scheme cannot be evenly distributed between the two acquisition units, and the acquisition unit connected to a single sensor cannot perform the phase similarity analysis. Similarity algorithms and warning outputs are also unsuitable. Secondly, from the perspective of reliability engineering and fault tolerance, the four accelerometers constitute a "2+2" redundant configuration. When any sensor fails, the system can automatically degrade to a three-sensor mode and continue to execute the complete discrimination process, thereby completely eliminating the risk of missed alarms caused by single-point failures. In contrast, the two-sensor scheme has no redundancy, and the failure of any one sensor will lead to system failure, which is unacceptable in major projects. In addition, the four-sensor scheme can greatly reduce the false alarm rate during normal operation through multi-dimensional criteria and time difference verification. In summary, the four-sensor scheme meets the minimum requirements of hardware logic and is also the technical choice that achieves the optimal balance between preventing missed alarms and preventing false alarms.

[0089] The present invention Figure 1 The fifth to thirteenth accelerometers serve as extended monitoring points, deployed on critical infrastructure or important building structures as needed for the project. Their core function is to collect seismic motion data at specific locations, without participating in real-time earthquake identification and early warning logic. The data they collect is dedicated to detailed post-earthquake data analysis and structural safety assessment, such as calculating cumulative absolute velocity (CAV) and generating response spectra, providing auxiliary basis for post-earthquake safety decisions. The number of accelerometers other than the first to fourth accelerometers is determined based on the distribution and scale of key facilities (number of important buildings or equipment) and structures in the specific project, as well as the required level of detail in the post-earthquake assessment. Figure 1 The fifth to thirteenth accelerometers shown in the figure represent only a typical configuration and do not constitute a limitation on the number of extended monitoring points (i.e., other accelerometers).

[0090] The present invention Figure 1The functions of the third to seventh strong-motion recorders are as follows: These recorders serve as extended data acquisition units, connected to the sensors at the aforementioned extended monitoring points; their main function is to acquire, cache, and forward the raw acceleration data from the connected sensors to the central processing unit 7 with high fidelity; they do not possess real-time earthquake discrimination algorithms, nor do they directly drive early warning outputs; their existence ensures that the system can achieve rapid and reliable early warning while completely recording the seismic motion time history data of all key points in the entire station, meeting the needs for in-depth post-event analysis and archiving; the number of strong-motion recorders other than the first to second strong-motion recorders is determined based on the number and location distribution of the acceleration sensors at the connected extended monitoring points. Figure 1 The strong-motion recorders 3 through 7 shown in the figure are merely examples and do not constitute a limitation on the number of extended data acquisition units (i.e., other strong-motion recorders).

[0091] In the first step of this invention, the uniform arrangement of the four accelerometers means that the distance between each pair of the four accelerometer placement points is equal or approximately equal, such as a square vertex layout. However, the core purpose is to achieve "spatial separation," that is, to ensure sufficient spacing to take advantage of the propagation differences between seismic waves and interference waves. Therefore, the arrangement focuses on meeting functional requirements rather than strict geometric uniformity. The four accelerometers are placed at the four corners of the engineering area, and although the spacing is not completely equal, effective spatial separation has been achieved, which fully meets the requirements of this invention.

[0092] In the third step of this invention, when performing waveform similarity calculations between the first and second processed signals, and between the third and fourth processed signals, the waveform data time period used is consistent with the waveform data time period used in the second step to extract local feature values ​​and record the early warning trigger time. Specifically, the data used for waveform similarity calculations for each processed signal are all taken from the waveform segment within the same time window corresponding to when the signal was determined to meet the earthquake feature threshold in the second step. This time window is a waveform data segment with a length of 1 second, starting from the arrival time of the P wave.

[0093] In the fourth step of this invention, the earthquake early warning parameters include the earthquake early warning magnitude and the epicenter location.

[0094] The earthquake early warning described in this invention refers to an early warning issued based on P-wave information before the arrival of the S-wave of an earthquake. After the earthquake early warning is issued, corresponding emergency measures can be activated, including: slowing down or stopping fast-moving trains to prevent derailment accidents; and actively disconnecting power supply to power facilities to avoid short circuits or fires caused by equipment damage.

[0095] The earthquake alarm described in this invention refers to the confirmation that the earthquake intensity has exceeded the preset safety limit after the arrival of the S-wave, indicating that the destructive seismic wave has reached the engineering area; at this time, emergency response measures should be implemented immediately, including forced emergency braking of trains and cutting off power supply, in order to minimize the losses caused by the earthquake disaster.

[0096] Example 1:

[0097] See Figures 1-3 A precise earthquake alarm system and method for major regional engineering projects based on four sensors is disclosed. The system includes four acceleration sensors, a first strong-motion recorder 5, a second strong-motion recorder 6, a central processing unit 7, and an alarm box 8. The four acceleration sensors are designated as first acceleration sensor 1, second acceleration sensor 2, third acceleration sensor 3, and fourth acceleration sensor 4. The first acceleration sensor 1 and second acceleration sensor 2 are connected to the first strong-motion recorder 5, and the third acceleration sensor 3 and fourth acceleration sensor 4 are connected to the second strong-motion recorder 6. The first strong-motion recorder 5 and the second strong-motion recorder 6 are respectively connected to the alarm box 8 via relays. Simultaneously, the first strong-motion recorder 5 and the second strong-motion recorder 6 are both communicatively connected to the central processing unit 7, and the central processing unit 7 is communicatively connected to the alarm box 8.

[0098] In application, the system constructs a direct hardware connection path and a software communication connection path based on the aforementioned physical connection to work together;

[0099] The direct hardware connection path is as follows: the first strong-motion recorder 5 and the second strong-motion recorder 6 are respectively connected to the alarm box 8 via relays; this path is dedicated to realizing rapid early warning in the four-sensor working mode: the first acceleration sensor 1, the second acceleration sensor 2, the third acceleration sensor 3, and the fourth acceleration sensor 4 respectively collect raw acceleration signals; among them, the raw acceleration signals of the first acceleration sensor 1 and the second acceleration sensor 2 are recorded in real time by the first strong-motion recorder 5 and preprocessed to obtain the first processed signal and the second processed signal; then, the first strong-motion recorder 5 performs feature extraction, intra-group waveform similarity, and time difference judgment on the first processed signal and the second processed signal; simultaneously, the raw acceleration signals of the third acceleration sensor 3 and the fourth acceleration sensor 4 are recorded in real time by the second strong-motion recorder 6 and preprocessed to obtain the third processed signal and the fourth processed signal; then, the second strong-motion recorder 6 performs feature extraction, intra-group waveform similarity, and time difference judgment on the third processed signal and the fourth processed signal. The system then performs feature extraction, waveform similarity analysis within groups, and time difference determination on the signals. When both the first strong-motion recorder 5 and the second strong-motion recorder 6 independently determine an earthquake event, they simultaneously send hardware trigger signals to the alarm box 8 via relays to issue an earthquake early warning. After issuing the earthquake early warning, the first and second strong-motion recorders 5 and 6 continue to monitor the processed signals from the connected acceleration sensors. When the first strong-motion recorder 5 determines that both the first and second processed signals exceed the preset earthquake alarm threshold, and the second strong-motion recorder 6 determines that both the third and fourth processed signals exceed the threshold, the two strong-motion recorders again send hardware trigger signals to the alarm box 8 via relays, driving the alarm box 8 to issue an earthquake alarm signal, indicating that the earthquake intensity has reached a hazardous level. This pure hardware path bypasses the software aspects of the central processing unit 7, enabling rapid hardware response from perception to early warning to alarm, thus improving the system's timeliness.

[0100] The software communication connection path is as follows: both the first strong-motion recorder 5 and the second strong-motion recorder 6 are connected to the central processing unit 7, and the central processing unit 7 is connected to the alarm box 8. This path enables system monitoring and fault tolerance: when the hardware path is working normally, the central processing unit 7 continuously receives data and performs monitoring; once any sensor failure is detected, the system immediately and automatically switches to the three-sensor working mode controlled by the central processing unit 7; at this time, the raw acceleration signals collected by the effective acceleration sensors in the first acceleration sensor 1 and the second acceleration sensor 2 are transmitted to the central processing unit 7 through the first strong-motion recorder 5, and the third acceleration sensor 3 and the fourth acceleration sensor 4... The raw acceleration signal collected by the effective accelerometer is transmitted to the central processing unit (CPU) 7 via the second strong-motion recorder 6. After preprocessing all signals, the CPU 7 executes a strictly adapted three-sensor discrimination logic based on the data from the remaining three normal sensors. If an earthquake event is determined, the CPU 7 sends an earthquake early warning command to the alarm box 8 to issue an earthquake early warning. After issuing the earthquake early warning, the CPU 7 continues to monitor the processed signals from the three effective sensors. If the processed signals from all effective sensors exceed the preset earthquake alarm threshold, the CPU 7 drives the alarm box 8 to issue an earthquake alarm signal, thereby eliminating the risk of missed alarms due to a single point of failure causing the complete loss of the early warning function.

[0101] Furthermore, when all four sensors are working normally, the system is in a standard operating mode. In this mode, the hardware direct connection path and the communication connection path operate simultaneously, while the system's final warning, alarm triggering, and output are completed independently by the hardware direct connection path.

[0102] Through the dual-path architecture based on clear physical distinctions, the system ensures the timeliness of early warning and alarm under normal conditions by relying on the hardware path, while ensuring the continuity of functions through intelligent degradation of the software path in case of failure, thus achieving a balance between high speed and high reliability.

[0103] Example 2:

[0104] The basic content is the same as in Example 1, except that the precise alarm method includes the following steps:

[0105] Step 1: Within the area covered by the major project, four accelerometers are evenly deployed; the four accelerometers collect data synchronously to obtain four raw acceleration signals, and each raw acceleration signal is preprocessed to obtain four processed signals, namely the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal.

[0106] Step 2: First, extract the local feature values ​​of the four processed signals respectively, and then compare the local feature values ​​of the four processed signals with the corresponding earthquake feature thresholds respectively; if the local feature values ​​of the four processed signals all meet the corresponding earthquake feature thresholds, then record an early warning trigger time for each processed signal and proceed to Step 3; otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0107] Step 3: First, perform waveform similarity calculations between the first and second processed signals, and between the third and fourth processed signals, to obtain the first similarity value and the second similarity value. Then, compare the first and second similarity values ​​with the seismic wave similarity judgment threshold. If both the first and second similarity values ​​meet the seismic wave similarity judgment threshold, proceed to step 4. Otherwise, it is determined that no earthquake event has occurred, and the early warning process terminates.

[0108] Step 4: First, based on the warning trigger time of each processed signal recorded in Step 2, calculate the first warning trigger time difference between the first and second processed signals, and the second warning trigger time difference between the third and fourth processed signals; then compare the first and second warning trigger time differences with the time difference range of seismic waves; if either the first or second warning trigger time difference does not conform to the time difference range of seismic waves, it is determined that no earthquake event has occurred, and the warning process terminates; if both the first and second warning trigger time differences conform to the time difference range of seismic waves, it is determined that an earthquake event has occurred, earthquake warning parameters are calculated, and an earthquake warning is issued.

[0109] Step 5: After issuing an earthquake early warning, continue to monitor the processed signals of the above-mentioned acceleration sensors; if the processed signal of any acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm will be issued and the alarm process will end; if the processed signals of all four acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm will be issued.

[0110] In application, in the first step, the four accelerometers synchronously acquire four raw acceleration signals, which are achieved by the first accelerometer 1, the second accelerometer 2, the third accelerometer 3, and the fourth accelerometer 4. The preprocessing of each raw acceleration signal is achieved by the first strong-motion recorder 5 and the second strong-motion recorder 6, respectively. The first strong-motion recorder 5 is responsible for the raw acceleration signals of the first accelerometer 1 and the second accelerometer 2, and the second strong-motion recorder 6 is responsible for the raw acceleration signals of the third accelerometer 3 and the fourth accelerometer 4.

[0111] In the second step, the extraction of local feature values ​​of the four processed signals and the comparison of the local feature values ​​of the four processed signals with the corresponding seismic feature thresholds are respectively implemented by the first strong motion recorder 5 and the second strong motion recorder 6.

[0112] In the third step, the waveform similarity calculation between the first processed signal and the second processed signal, and the comparison with the seismic wave similarity judgment threshold, are implemented by the first strong motion recorder 5; the waveform similarity calculation between the third processed signal and the fourth processed signal, and the comparison with the seismic wave similarity judgment threshold, are implemented by the second strong motion recorder 6.

[0113] In the fourth step, the first warning trigger time difference between the calculated first processed signal and the second processed signal, and the comparison with the time difference range of the seismic wave, are implemented by the first strong motion recorder 5; the second warning trigger time difference between the third processed signal and the fourth processed signal, and the comparison with the time difference range of the seismic wave, are implemented by the second strong motion recorder 6; the alarm box 8 is configured to execute the final earthquake warning output when valid trigger signals from two strong motion recorders are received simultaneously.

[0114] After issuing an earthquake warning, the first strong-motion recorder 5 and the second strong-motion recorder 6 calculate earthquake warning parameters using the P-wave and drive the alarm box 8 to issue an earthquake warning signal. Subsequently, the system waits for a preset time (e.g., 40 seconds) to monitor subsequent seismic waves. During this period, the four acceleration sensors continuously collect acceleration values. The first strong-motion recorder 5 determines whether the processed signals (pre-processed acceleration values) of the first and second acceleration sensors exceed the preset earthquake alarm threshold (i.e., safety limit). The second strong-motion recorder 6 determines whether the processed signals of the third and fourth acceleration sensors exceed the threshold. If the acceleration value of any one sensor does not exceed the threshold, the earthquake alarm is not triggered. If the acceleration values ​​of all four sensors exceed the threshold, the two strong-motion recorders send hardware trigger signals to the alarm box 8 via relays, driving the alarm box 8 to issue an earthquake alarm signal, indicating that the vibration intensity of this earthquake event has exceeded the safety limit and reached a hazardous level.

[0115] The first to fourth steps constitute the earthquake early warning stage, the main purpose of which is to accurately identify earthquake events through multi-level criteria (local characteristics, waveform similarity, time difference range), providing a precise and reliable triggering basis for subsequent alarms. The fifth step is the earthquake alarm stage, the main purpose of which is to promptly issue an alarm signal after the earthquake early warning by monitoring whether the amplitude of the processed signal exceeds the earthquake alarm threshold and confirming that the ground motion intensity has reached a hazardous level. The early warning stage provides accurate event identification and early warning triggering for the alarm stage, while the alarm stage further confirms the disaster intensity based on the early warning. Together, they achieve a complete closed loop from earthquake event detection to emergency response.

[0116] Example 3:

[0117] The basic content is the same as in Example 1, except that the precise alarm method also includes a fault-tolerant mode.

[0118] The system continuously monitors the signals from the four aforementioned accelerometer sensors. If any one of the accelerometer signals fails, it automatically switches to a three-sensor operating mode based on the remaining three valid sensor signals, while maintaining the first step described above. The three-sensor operating mode includes:

[0119] In the second step, the local feature values ​​of the three processed signals are extracted first, and then the local feature values ​​of the three processed signals are compared with the corresponding earthquake feature thresholds. If the local feature values ​​of the three processed signals all meet the corresponding earthquake feature thresholds, an early warning trigger time is recorded for each processed signal, and the process proceeds to the third step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0120] In the third step, the waveform similarity between each pair of the three processed signals is calculated to obtain the waveform similarity value between each pair. Then, the waveform similarity value between each pair is compared with the seismic wave similarity judgment threshold. If all the similarity values ​​between each pair meet the seismic wave similarity judgment threshold, the process proceeds to the fourth step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated.

[0121] In the fourth step, based on the warning trigger time of each processed signal recorded in the second step, the warning trigger time difference between each pair of the three processed signals is calculated. Then, all the warning trigger time differences between each pair are compared with the time difference range of the seismic waves. If any one of the warning trigger time differences between each pair does not conform to the time difference range of the seismic waves, it is determined that no earthquake event has occurred, and the warning process is terminated. If all the warning trigger time differences between each pair conform to the time difference range of the seismic waves, it is determined that an earthquake event has occurred, and earthquake warning parameters are calculated and an earthquake warning is issued.

[0122] In the fifth step, after issuing an earthquake early warning, the processed signals of the three acceleration sensors continue to be monitored; if the processed signal of any one acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm is issued and the alarm process ends; if the processed signals of all three acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm is issued.

[0123] When any sensor fails, the system can automatically and seamlessly switch to a three-sensor operating mode and continue operating based on the reconfigured discrimination logic (including local feature comparison of the three valid signals, pairwise waveform similarity verification, and early warning time difference verification). This fundamentally eliminates the risk of complete loss of earthquake early warning function due to single-point sensor failure, ensures the continuity of earthquake early warning, significantly reduces the false alarm rate, and greatly improves the reliability and practicality of earthquake early warning.

[0124] In the three-sensor operating mode, the central processing unit acquires data from the remaining normal sensors through a communication connection and takes over the core discrimination function of the system, independently completing all processing and judgment logic in steps two through four; if it is determined to be an earthquake event, the central processing unit 7 sends a warning command to the alarm box 8 to drive the warning output;

[0125] After issuing an earthquake warning, the central processing unit 7 calculates earthquake warning parameters using the P-wave and drives the alarm box 8 to issue an earthquake warning signal. Subsequently, the system waits for a preset time (e.g., 40 seconds) to monitor subsequent seismic waves. During this period, three effective accelerometers continuously collect acceleration values, and the central processing unit 7 determines whether the processed signals (pre-processed acceleration values) of each effective accelerometer exceed the preset earthquake alarm threshold (i.e., safety limit). If the acceleration value of any one sensor does not exceed the threshold, the earthquake alarm is not triggered. If the acceleration values ​​of all effective accelerometers exceed the threshold, the central processing unit 7 drives the alarm box 8 to issue an earthquake alarm signal, indicating that the vibration intensity of this earthquake event has exceeded the safety limit and reached a hazardous level.

[0126] The first to fourth steps constitute the earthquake early warning stage. Their main purpose is to accurately identify earthquake events using multi-level criteria (local characteristics, waveform similarity, time difference range), providing a precise and reliable triggering basis for subsequent alarms. The fifth step is the earthquake alarm stage. Its main purpose is to promptly issue an alarm signal after the earthquake early warning, by monitoring whether the amplitude of the processed signal exceeds the earthquake alarm threshold and confirming that the ground motion intensity has reached a hazardous level. The early warning stage provides accurate event identification and early warning triggering for the alarm stage, while the alarm stage further confirms the disaster intensity based on the early warning. Together, they achieve a complete closed loop from earthquake event detection to emergency response.

[0127] Furthermore, once the sensor malfunction is repaired, the system automatically identifies that the sensor signal has returned to normal by real-time detection of the signals from the four accelerometers; then it automatically and seamlessly switches back to the standard four-sensor operating mode and fully restores all the discrimination logic and processing flow described in Example 2.

[0128] Example 4:

[0129] The basic content is the same as in Embodiment 1, except that in the first step, the four acceleration sensors are placed in the free field, and each acceleration sensor is installed through an instrument pier, which is connected to the bedrock or old soil layer below the free field.

[0130] In application, four sensors are first evenly distributed in the area covered by the major project, and the distance between the four sensors is maximized. This reduces the similarity of the waveforms when vibrations generated by various local interference sources such as machinery and vehicles inside the project site are transmitted to different sensors, thereby enhancing the ability to distinguish between local interference and real earthquake events from the perspective of physical layout.

[0131] Secondly, the sensors are deployed in a free field. When selecting a location in a free field, the sensors actively avoid tall buildings and large equipment foundations. This avoids interference sources that may block, reflect, or cause additional vibrations to the propagation of seismic waves, thus ensuring the spatial representativeness of the collected signals.

[0132] Finally, each sensor is installed using an instrument pier, with the installation depth ensuring firm contact with the underlying bedrock or old soil layer. This eliminates the differences caused by the "site amplification effect" due to the uneven thickness or properties of the soft soil layer on the site surface, ensuring that the four sensors are established on a unified and stable vibration reference plane. This provides a consistent, comparable, and accurate data foundation for the subsequent analysis of amplitude and waveform characteristics of multi-sensor data, avoiding recording distortion caused by differences in local site conditions.

[0133] Example 5:

[0134] The basic content is the same as in Example 1, except that: in the first step, the preprocessing includes noise reduction processing; the noise reduction processing is: performing 0.1Hz-33Hz bandpass filtering on each original acceleration signal.

[0135] In application, the effective energy of seismic waves (especially their initial P-waves and the main energy S-waves and surface waves) is mainly distributed in the frequency band below tens of hertz; setting the lower bandpass limit to 0.1 Hz can preserve these key low-frequency components that reflect the nature of seismic events, while effectively filtering out extremely low-frequency drift caused by the instrument itself and the environment.

[0136] Secondly, the bandpass limit is set to 33Hz, mainly for the following two reasons:

[0137] Firstly, suppressing high-frequency environmental noise: High-frequency (usually above 33Hz) environmental vibrations and transient interferences are commonly generated at engineering sites by people walking, vehicles passing, and small motors running; these noise components are significantly different from seismic waves in terms of spectrum, but their superposition will seriously reduce the signal-to-noise ratio and interfere with subsequent feature extraction and waveform similarity analysis; setting an upper limit of 33Hz can effectively filter out such irrelevant high-frequency noise.

[0138] Secondly, retaining key ground motion information for safety assessment of major projects: For major projects such as nuclear power plants and precision instrument factories, electrical equipment and sensitive instruments installed on floors or structures are highly sensitive to higher frequency ground motions (usually covering tens of hertz). 33Hz is often a key frequency control point in vibration control and seismic analysis of such structures, and vibration energy related to this frequency band is closely related to seismic damage to electrical equipment. Therefore, retaining ground motion frequency components below 33 Hz in preprocessing not only for earthquake event identification but also provides the necessary raw data foundation for subsequent seismic damage assessments or special analyses that may be triggered and are directly related to project safety.

[0139] Therefore, by using bandpass filtering in the 0.1Hz-33Hz range, the original acceleration signal was initially purified during the preprocessing stage. This effectively preserved the core frequency components of the seismic waves and the higher-frequency dynamic information crucial for the safety of major engineering projects, while maximally eliminating irrelevant extremely low-frequency drift and high-frequency environmental noise. This provides a high-quality, high-signal-to-noise ratio input signal for subsequent intelligent discrimination processes such as accurate calculation of local features and reliable waveform similarity comparison. From a data preprocessing perspective, this ensures the effectiveness and reliability of the entire system's identification algorithm while also maintaining the integrity of the engineering safety data.

[0140] Example 6:

[0141] The basic content is the same as in Example 1, except that in the second step, extracting the local feature values ​​of the four processed signals and comparing them with the corresponding seismic feature thresholds means:

[0142] Perform the following operations on the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal, respectively:

[0143] First, the STA / LTA algorithm is used to coarsely pick up the first arrival of the P-wave in the processed signal to obtain the coarse arrival time; then, the VAR-AIC algorithm is used to accurately pick up the first arrival of the P-wave within a preset time window, using the coarse arrival time as a reference point, to determine the first arrival time of the P-wave.

[0144] Starting from the initial arrival time of the P wave, waveform data for 1 second is extracted. Then, based on this waveform data, the number of zero crossings, kurtosis growth factor, and spectral energy ratio of the processed signal are calculated.

[0145] Finally, the number of zero-crossings of each processed signal is compared with the seismic feature threshold corresponding to the number of zero-crossings, the kurtosis growth coefficient of each processed signal is compared with the seismic feature threshold corresponding to the kurtosis growth coefficient, and the spectral energy ratio of each processed signal is compared with the seismic feature threshold corresponding to the spectral energy ratio.

[0146] In application, reliable initial screening of seismic events is achieved at the single-sensor level through P-wave picking and joint criteria of multiple features, as detailed below:

[0147] For P-wave picking: The STA / LTA algorithm is used to quickly locate the approximate location of the first arrival of the P-wave. Then, with this location as the center, the VAR-AIC algorithm is used to perform a high-resolution search within a local time window to obtain the precise time of the first arrival of the P-wave. This method can stably pick up the first arrival of the P-wave in a low signal-to-noise ratio environment, providing a time-aligned waveform analysis window for subsequent feature value calculation, and avoiding calculation deviations caused by time window offset.

[0148] The STA / LTA algorithm (short-time average to long-time average ratio method) sets short-time windows and long-time windows on the signal time series, calculates the average value of the signal characteristic function within the short-time window and the average value of the signal characteristic function within the long-time window, and calculates the ratio between the two; when the ratio exceeds a preset threshold, it is determined as the approximate arrival time of the P wave.

[0149] The VAR-AIC algorithm (Akaike Information Criterion Based on Variance) is a high-precision phase picking method. Its core approach is to take the seismic signal sequence within the time window to be analyzed as input, calculate the variance of the sequence into two segments by sliding point by point, and construct an AIC function based on the variance. The minimum point of the AIC function is the position where the change in the statistical characteristics of the signal is most significant, which corresponds to the precise time of the first arrival of the P wave. This method does not require the estimation of autoregressive coefficients, has high computational efficiency, and still has good picking accuracy under low signal-to-noise ratio conditions.

[0150] The processed signals from each sensor are processed in parallel, and the following three local features are calculated and judged to form a multi-dimensional intelligent initial screening unit:

[0151] Zero-crossing count: The number of times a signal crosses the zero axis per unit time is used to characterize the dominant frequency of the signal. Since the initial energy of seismic P-waves is concentrated in the low-frequency band, the zero-crossing count is relatively low. However, the dominant frequency of high-frequency interference in engineering sites (such as mechanical operation and personnel activities) is usually higher than 30Hz, resulting in a significantly higher zero-crossing count. Therefore, the zero-crossing count can be used to filter out vibrations in such high-frequency environments.

[0152] Kurodivo Growth Coefficient (PGC): This is the absolute maximum value of the acceleration difference between adjacent sampling points, used to quantify the intensity of instantaneous changes in the signal. The initial motion of a seismic P-wave is characterized by a slow increase in amplitude, resulting in a small PGC value. However, impact-type disturbances (such as blasting and impact) cause sudden acceleration changes within milliseconds, leading to a sharp increase in the PGC value. PGC can be used to identify and eliminate these types of instantaneous strong impact events that are most likely to cause false alarms using traditional methods.

[0153] Spectral Energy Ratio (SER) discrimination: Calculate the energy ratio of a specified low-frequency band (e.g., 0.1-5Hz) to a high-frequency band (e.g., 20-33Hz); seismic waves have predominantly low-frequency energy, and the SER value is usually greater than 1; most artificial vibration sources have energy concentrated in the high-frequency band, and the SER value is less than 1; by setting a lower limit threshold for the ratio (e.g., >1), seismic signals and mechanical vibrations can be distinguished from each other in terms of energy distribution.

[0154] By combining the above-mentioned time-domain (zero-crossing number, PGC) and frequency-domain (SER) characteristics, the multi-dimensional essential characteristics of vibration events can be examined simultaneously at the single sensor level. Only signals that simultaneously meet all three criteria are initially identified as having potential seismic characteristics and enter the subsequent multi-sensor collaborative analysis process. This improves signal quality and initial screening reliability from the source of algorithm logic, laying the foundation for achieving a low false alarm rate.

[0155] Example 7:

[0156] The basic content is the same as in Example 1, except that the seismic feature threshold corresponding to the number of zero crossings is a preset upper frequency value; when the number of zero crossings of the processed signal is not greater than the upper frequency value, the processed signal meets the seismic feature threshold corresponding to the number of zero crossings.

[0157] In application, the preset upper frequency limit is the zero-crossing number threshold used to distinguish between seismic waves and engineering interference. This threshold is based on the fundamental differences in the spectral characteristics of seismic waves and engineering interference: the dominant frequency of destructive seismic waves is mainly distributed in the low-frequency range of 0.1Hz-10Hz, with a low zero-crossing number; while the vibration energy generated by interference sources at the engineering site (such as machinery, vehicle traffic, construction blasting, etc.) is significantly concentrated in the 10Hz-100Hz or even higher frequency band, with a high zero-crossing number. Therefore, setting the upper frequency limit to a level that can effectively distinguish the statistical difference in zero-crossing numbers between typical seismic P-wave signals and typical high-frequency environmental interference can effectively suppress the influence of high-frequency environmental vibrations during the feature extraction stage.

[0158] Specifically, the initial upper limit of the frequency is set to no more than 55 zero-crossings of the signal within 1 second. This value is an engineering balance point determined by combining historical earthquake P-wave zero-crossing statistics (generally <50 times / second) and on-site strong interference zero-crossing statistics (usually >80 times / second), and optimized through anti-false alarm simulation tests. Furthermore, in practical applications, the seismic characteristic threshold corresponding to the number of signal zero-crossings is verified and fine-tuned based on the monitoring data and event records of the specific site.

[0159] Example 8:

[0160] The basic content is the same as in Example 1, except that: the seismic feature threshold corresponding to the kurtosis growth coefficient is a preset upper limit of the rate of change; when the kurtosis growth coefficient of the processed signal is not greater than the upper limit of the rate of change, the processed signal meets the seismic feature threshold corresponding to the kurtosis growth coefficient.

[0161] In application, the preset upper limit of the rate of change refers to the critical value of the kurtosis growth coefficient used to distinguish between seismic waves and engineering interference. The basis for setting it is the essential difference between seismic waves and engineering interference in the instantaneous change characteristics of the signals: due to the physical constraints of the source rupture and propagation process, the acceleration amplitude of seismic P waves shows a relatively slow and continuous climbing process, with a limited rate of change between adjacent sampling points and a low kurtosis growth coefficient; while events such as blasting, impact, and falling heavy objects at the engineering site release energy violently within milliseconds, causing the acceleration signal to undergo step or spike-like abrupt changes, with a very large instantaneous rate of change and a high kurtosis growth coefficient;

[0162] The kurtosis growth factor (PGC) is a direct time-domain parameter that quantifies the intensity of instantaneous changes in a signal. Based on the essential differences in the physical mechanisms mentioned above, the upper limit of the rate of change is set at a level that can effectively distinguish between typical earthquake P-wave initial motion signals and typical instantaneous impact interference. This can effectively identify and eliminate the instantaneous strong interference that is most likely to cause false alarms during the feature extraction stage, thereby improving the reliability of single-point signal identification.

[0163] The kurtosis growth factor (PGC) is the absolute maximum value of the acceleration difference between adjacent sampling points, and its calculation formula is: ;

[0164] in, and This represents the acceleration values ​​at two consecutive sampling points, expressed in gal (gal).

[0165] Specifically, the upper limit of the rate of change is initially set to a kurtosis growth coefficient of no more than 40 gal. This value is the engineering equilibrium point determined by comparing and analyzing the statistical characteristics of the P-wave initial motion segment in historical earthquake records (whose PGC values ​​are mostly distributed in the range of 10-30 gal) with typical impact events measured at the target engineering site (such as blasting and impact, whose PGC values ​​generally exceed 80 gal), and by simulation testing and optimization. Furthermore, in practical applications, the seismic characteristic threshold corresponding to the kurtosis growth coefficient is verified and fine-tuned based on the monitoring data and event records of the specific site.

[0166] Example 9:

[0167] The basic content is the same as in Example 1, except that the seismic feature threshold corresponding to the spectral energy ratio is a preset lower limit value; when the spectral energy ratio of the processed signal is not less than the lower limit value, the processed signal meets the seismic feature threshold corresponding to the spectral energy ratio.

[0168] In application, the preset lower limit of the ratio refers to the critical value of the spectral energy ratio used to distinguish between seismic waves and engineering interference. It is set based on the essential difference in the energy spectrum distribution between seismic waves and engineering interference. The spectral energy ratio is the energy ratio of the low-frequency band and the high-frequency band, where the low-frequency band is below 5Hz and the high-frequency band is above 20Hz. The large rupture scale of the hypocenter of a destructive earthquake results in long dominant wavelengths of the seismic waves it generates, and the high-frequency components attenuate faster during propagation. As a result, the signal energy reaching the Earth's surface is mainly concentrated in the low-frequency band (below 5Hz), and the spectral energy ratio is relatively high. On the other hand, artificial vibration sources such as mechanical equipment, vehicle operation, and impacts have rich excitation frequencies or high-frequency resonant components, and the main energy is distributed in the mid-to-high frequency band (above 20Hz), resulting in a relatively low spectral energy ratio.

[0169] Based on the physical fact that low-frequency energy dominates in seismic waves, the lower limit of the ratio is set at a level that can effectively distinguish typical seismic signals (low-frequency energy dominant) from typical engineering interference (high-frequency energy dominant). This allows for filtering based on the energy distribution of the signal during the feature extraction stage, improving the specificity of seismic event identification. The spectral energy ratio is used to quantify the distribution comparison of signal energy in high and low frequency bands; its calculation formula is as follows: ;

[0170] in, and These are the acceleration values ​​of the low-frequency component (e.g., 0.1-5Hz) and high-frequency component (e.g., 20-33Hz) at the i-th sampling point, respectively, obtained after the processed signal has been filtered by a specific bandpass filter.

[0171] Specifically, the initial lower limit of the ratio is set to a spectral energy ratio greater than 1. This value is determined by comparing the analysis of historical strong earthquake records (the SER values ​​of the P-wave and S-wave initial motion segments calculated according to this definition are usually significantly greater than 1) with the measured analysis of various typical disturbances at the engineering site (the SER values ​​of most of them are less than 1), and is verified by simulation testing as an engineering equilibrium point. Furthermore, in practical applications, the seismic characteristic threshold corresponding to the spectral energy ratio is verified and optimized based on the monitoring data and event records of the specific site.

[0172] Example 10:

[0173] The basic content is the same as in Example 1, except that in the third step, the waveform similarity is obtained by calculating the cross-correlation coefficient of the two processed signals; the seismic wave similarity judgment threshold is set according to the high correlation between the waveforms recorded at different spatial locations of real earthquake events.

[0174] In application, the seismic wave similarity judgment threshold refers to the critical value of the cross-correlation coefficient used to distinguish between seismic waves and engineering interference. It is set based on the essential difference in spatial correlation between seismic waves and engineering interference. Real earthquake events originate from the same underground source. The seismic waves (especially the first arrival P waves) generated by these events, when propagating to spatially separated sensors, although there will be slight changes in wave arrival time, amplitude, and phase due to differences in propagation paths, maintain a high degree of homogeneity in the basic waveform, vibration sequence, and phase relationship, resulting in a high cross-correlation coefficient. This homogeneity is guaranteed by a common source mechanism, a relatively homogeneous propagation medium, and the determinism of the wave equation. In contrast, local interference at the engineering site (such as mechanical impact, vehicle traffic, blasting, etc.) usually has strong locality and directionality in space. After its vibration propagates through complex surface paths, reverses, superimposes, and rapidly attenuates, the waveform morphology recorded on sensors at different locations differs significantly, and the cross-correlation coefficient is generally low.

[0175] To quantify waveform similarity, cross-correlation coefficients are used. As an evaluation metric, for any two processed acceleration signals x and y from different sensors, the cross-correlation coefficient is calculated using the following formula within a set analysis time window:

[0176] ;

[0177] Where cov(x,y) is the covariance of signals x and y, and D(x) and D(y) are the variances of signals x and y, respectively. The value range is [-1, 1]. The closer the value is to 1, the more similar the two waveforms are in shape.

[0178] Based on the essential differences in spatial correlation mentioned above, setting the seismic wave similarity judgment threshold to a level that can effectively distinguish between typical seismic signals (highly correlated) and typical engineering interference (lowly correlated) can effectively eliminate local interference that does not have spatial correlation and improve the reliability of seismic event identification.

[0179] Specifically, the initial threshold for seismic wave similarity judgment is set to the cross-correlation coefficient. >0.65; This value is determined by comparing the cross-correlation coefficients of waveforms recorded by adjacent sensors at the hundreds of meters to kilometers level within a suitable P-wave initiation time window with the results of field measurements and simulation tests in engineering projects (the cross-correlation coefficients of waveforms for most local interference events on different sensors are usually below 0.3); The threshold of >0.65 provides sufficient margin for reasonable waveform differences caused by local variations in the site and noise superposition, while effectively eliminating most local interferences that do not have spatial correlation, thus ensuring low false alarms while ensuring low missed alarms. Furthermore, in practical applications, the seismic wave similarity judgment threshold is verified and fine-tuned based on the monitoring data and event records of specific sites.

[0180] Example 11:

[0181] The basic content is the same as in Example 1, except that in the fourth step, the time difference range of the seismic wave is the theoretical wave arrival time difference range calculated based on the preset P-wave propagation velocity, S-wave propagation velocity and the actual distance between each acceleration sensor.

[0182] In application, the time difference range of seismic waves refers to the critical range of the warning trigger time difference used to distinguish between seismic waves and engineering interference. It is set based on the essential difference in propagation speed between seismic waves and local surface interference. The propagation speed of seismic P waves and S waves in the crustal medium (usually on the order of several km / s) is much higher than the ground vibration propagation speed caused by common local interferences at engineering sites (such as vehicle traffic, mechanical vibration, construction blasting, etc.). The latter mainly relies on the surface soil layer for propagation, and the speed is significantly lower than that of seismic body waves. Therefore, by checking whether the warning trigger time difference of each acceleration sensor is within a reasonable time window calculated based on the physical laws of seismic wave propagation, it is possible to fundamentally distinguish between seismic events originating from distant sources and near-field surface interference from the perspective of propagation mechanism.

[0183] Based on the fundamental differences in propagation speed mentioned above, the average wave velocity of seismic waves in typical regions is used as the calculation benchmark: P-wave velocity The speed is 6.5 km / s, the speed of an S-wave. The actual distance between the two sensors after the sensors are deployed is 3.5 km / s. (Unit: km) Determined by measurement, and usually does not exceed 1 km (the area of ​​major projects generally does not exceed 3 km).

[0184] Assuming the earthquake event is located outside the monitoring area, the seismic waves (taking the first arriving P wave as an example) travel at approximately a constant velocity. Propagation; then the theoretical time difference between the arrival of the seismic wave at the i-th sensor and the reference sensor. Similarly, if the trigger wave is an S-wave, its theoretical arrival time difference is: Since the S-wave has a lower velocity, its theoretical arrival time difference is larger. Therefore, the theoretical arrival time difference of the S-wave should be used as a conservative criterion when making a judgment.

[0185] In the worst-case scenario (sensor distance to reference sensor) Calculated for 1km):

[0186] P-wave theory maximum time difference: ;

[0187] S-wave theory maximum time difference: ;

[0188] Therefore, the theoretical arrival time difference of seismic waves between sensors should not exceed 0.286s; considering the possible variations in wave velocity models and the uncertainty of source depth and orientation in actual geological conditions, an engineering safety margin needs to be added on the basis of the theoretical value.

[0189] Specifically, the time difference range of the seismic waves is initially set to 0.5s. This threshold covers the theoretical maximum arrival time difference (0.286s) while retaining sufficient margin to accommodate actual wave velocity fluctuations. It also forms a clear boundary with the typical arrival time difference pattern of local surface disturbances (which is often much greater than 0.5s). Furthermore, in practical applications, verification and adaptive adjustments are made based on monitoring data and event records from specific sites.

[0190] Example 12:

[0191] The basic content is the same as in Embodiment 1, except that: the first strong earthquake recorder 5, the second strong earthquake recorder 6 and the central processing unit 7 communicate with each other via TCP / IP protocol, and the central processing unit 7 communicates with the alarm box 8 via serial port.

[0192] In application, the TCP / IP connection enables the central processing unit 7 to collect and verify data from the first strong earthquake recorder 5 and the second strong earthquake recorder 6 in real time, providing support for fault-tolerant analysis of the software path; the serial port connection ensures that the central processing unit 7 can reliably send early warning commands to the alarm box 8 in fault-tolerant mode; this design achieves redundant verification through hardware and software collaboration; in terms of hardware, the relay outputs of the two sets of strong earthquake recorders serve as backups for each other, and triggering either set can drive the early warning, reducing the risk of missed alarms; in terms of software, the central processing unit 7 continuously monitors via TCP / IP and performs background fault-tolerant verification for each early warning; the dual communication and output configuration significantly improve the overall reliability of the system, ensuring that the early warning function can be executed continuously and reliably under any operating conditions.

[0193] The above description is only a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. Any equivalent modifications or changes made by those skilled in the art based on the content disclosed in the present invention should be included within the scope of protection set forth in the claims.

Claims

1. A precise earthquake early warning system for major regional engineering projects based on four sensors, characterized in that: The system includes four acceleration sensors, a first strong earthquake recorder (5), a second strong earthquake recorder (6), a central processing unit (7), and an alarm box (8); The four acceleration sensors are a first acceleration sensor (1), a second acceleration sensor (2), a third acceleration sensor (3), and a fourth acceleration sensor (4). The first acceleration sensor (1) and the second acceleration sensor (2) are connected to the first strong earthquake recorder (5), and the third acceleration sensor (3) and the fourth acceleration sensor (4) are connected to the second strong earthquake recorder (6). The first strong earthquake recorder (5) and the second strong earthquake recorder (6) are respectively connected to the alarm box (8) via relays; Meanwhile, the first strong earthquake recorder (5) and the second strong earthquake recorder (6) are both connected to the central processing unit (7), and the central processing unit (7) is connected to the alarm box (8).

2. A precise alarm method for a regional major engineering earthquake precision alarm system based on four sensors as described in claim 1, characterized in that: The precise alarm method includes the following steps: Step 1: Within the area covered by the major project, four accelerometers are evenly deployed; the four accelerometers collect data synchronously to obtain four raw acceleration signals, and each raw acceleration signal is preprocessed to obtain four processed signals, namely the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal. Step 2: First, extract the local feature values ​​of the four processed signals respectively, and then compare the local feature values ​​of the four processed signals with the corresponding earthquake feature thresholds respectively; if the local feature values ​​of the four processed signals all meet the corresponding earthquake feature thresholds, then record an early warning trigger time for each processed signal and proceed to Step 3; otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated. Step 3: First, perform waveform similarity calculations between the first and second processed signals, and between the third and fourth processed signals, to obtain the first similarity value and the second similarity value. Then, compare the first and second similarity values ​​with the seismic wave similarity judgment threshold. If both the first and second similarity values ​​meet the seismic wave similarity judgment threshold, proceed to step 4. Otherwise, it is determined that no earthquake event has occurred, and the early warning process terminates. Step 4: First, based on the warning trigger time of each processed signal recorded in Step 2, calculate the first warning trigger time difference between the first and second processed signals, and the second warning trigger time difference between the third and fourth processed signals; then compare the first and second warning trigger time differences with the time difference range of seismic waves; if either the first or second warning trigger time difference does not conform to the time difference range of seismic waves, it is determined that no earthquake event has occurred, and the warning process terminates; if both the first and second warning trigger time differences conform to the time difference range of seismic waves, it is determined that an earthquake event has occurred, earthquake warning parameters are calculated, and an earthquake warning is issued. Step 5: After issuing an earthquake early warning, continue to monitor the processed signals of the above-mentioned acceleration sensors; if the processed signal of any acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm will be issued and the alarm process will end; if the processed signals of all four acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm will be issued.

3. The method for accurate earthquake early warning in regional major engineering projects based on four sensors according to claim 2, characterized in that: The precise alarm method also includes a fault-tolerant mode: The system continuously monitors the signals from the four aforementioned accelerometer sensors. If any one of the accelerometer signals fails, it automatically switches to a three-sensor operating mode based on the remaining three valid sensor signals, while maintaining the first step described above. The three-sensor operating mode includes: In the second step, the local feature values ​​of the three processed signals are extracted first, and then the local feature values ​​of the three processed signals are compared with the corresponding earthquake feature thresholds. If the local feature values ​​of the three processed signals all meet the corresponding earthquake feature thresholds, an early warning trigger time is recorded for each processed signal, and the process proceeds to the third step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated. In the third step, the waveform similarity between each pair of the three processed signals is calculated to obtain the waveform similarity value between each pair. Then, the waveform similarity value between each pair is compared with the seismic wave similarity judgment threshold. If all the similarity values ​​between each pair meet the seismic wave similarity judgment threshold, the process proceeds to the fourth step. Otherwise, it is determined that no earthquake event has occurred, and the early warning process is terminated. In the fourth step, based on the warning trigger time of each processed signal recorded in the second step, the warning trigger time difference between each pair of the three processed signals is calculated. Then, all the warning trigger time differences between each pair are compared with the time difference range of the seismic waves. If any one of the warning trigger time differences between each pair does not conform to the time difference range of the seismic waves, it is determined that no earthquake event has occurred, and the warning process is terminated. If all the warning trigger time differences between each pair conform to the time difference range of the seismic waves, it is determined that an earthquake event has occurred, and earthquake warning parameters are calculated and an earthquake warning is issued. In the fifth step, after issuing an earthquake early warning, the processed signals of the three acceleration sensors continue to be monitored; if the processed signal of any one acceleration sensor does not exceed the earthquake alarm threshold, no earthquake alarm is issued and the alarm process ends; if the processed signals of all three acceleration sensors exceed the earthquake alarm threshold, an earthquake alarm is issued.

4. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 2 or 3, characterized in that: In the first step, the four acceleration sensors are deployed in the free field. Each acceleration sensor is installed via an instrument pier, which is connected to the bedrock or old soil layer below the free field.

5. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 2 or 3, characterized in that: In the first step, the preprocessing includes noise reduction processing; The noise reduction process involves performing a 0.1Hz-33Hz bandpass filter on each original acceleration signal.

6. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 2 or 3, characterized in that: In the second step, extracting the local feature values ​​of the four processed signals respectively, and then comparing the local feature values ​​of the four processed signals with the corresponding seismic feature thresholds, means: Perform the following operations on the first processed signal, the second processed signal, the third processed signal, and the fourth processed signal, respectively: First, the STA / LTA algorithm is used to coarsely pick up the first arrival of the P-wave in the processed signal to obtain the coarse arrival time; then, the VAR-AIC algorithm is used to accurately pick up the first arrival of the P-wave within a preset time window, using the coarse arrival time as a reference point, to determine the first arrival time of the P-wave. Starting from the initial arrival time of the P wave, waveform data for a duration of 1 second is extracted. Then, based on the waveform data, the number of zero crossings, kurtosis growth factor, and spectral energy ratio of the processed signal are calculated. Finally, the number of zero-crossings of each processed signal is compared with the seismic feature threshold corresponding to the number of zero-crossings, the kurtosis growth coefficient of each processed signal is compared with the seismic feature threshold corresponding to the kurtosis growth coefficient, and the spectral energy ratio of each processed signal is compared with the seismic feature threshold corresponding to the spectral energy ratio.

7. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 6, characterized in that: The seismic feature threshold corresponding to the number of zero crossings is a preset upper frequency value; when the number of zero crossings of the processed signal is not greater than the upper frequency value, the processed signal meets the seismic feature threshold corresponding to the number of zero crossings.

8. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 6, characterized in that: The seismic characteristic threshold corresponding to the kurtosis growth coefficient is a preset upper limit of the rate of change; when the kurtosis growth coefficient of the processed signal is not greater than the upper limit of the rate of change, the processed signal meets the seismic characteristic threshold corresponding to the kurtosis growth coefficient.

9. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 6, characterized in that: The seismic characteristic threshold corresponding to the spectral energy ratio is a preset lower limit value; when the spectral energy ratio of the processed signal is not less than the lower limit value, the processed signal meets the seismic characteristic threshold corresponding to the spectral energy ratio.

10. A method for accurate earthquake early warning in regional major engineering projects based on four sensors, as described in claim 2 or 3, characterized in that: In the fourth step, the time difference range of the seismic wave is the theoretical wave arrival time difference range calculated based on the preset P-wave propagation velocity, S-wave propagation velocity and the actual distance between each acceleration sensor.

Images