A method and system for seismic geomagnetic observation data acquisition transmission

Abstract

The application discloses a kind of for seismic geomagnetic observation data acquisition transmission method and system, it is related to the field of earthquake monitoring technology, including, acquisition band event marker original waveform data, based on impulse neural network computing environment executes coordinated impulse emission, obtains enhanced pulse timing sequence;Using the pulse filtering algorithm based on dynamic amplitude threshold, in combination with phase-locked loop synchronization method optimization enhanced pulse timing sequence Information density and timing consistency, generate high signal-to-noise ratio optimization pulse sequence;Optimized pulse sequence is packaged as standard transmission frame and is transmitted to seismic detection center, while decoding reconstruction is carried out through CNN-LSTM model, generate high-fidelity seismic geomagnetic observation data.Through CNN-LSTM model, using CNN convolution layer extracts the spatial characteristics of transmission frame, in combination with LSTM network learning timing dependence, realizes the accurate reconstruction of compressed transmission data.

Description

Technical Field

[0001] This invention relates to the field of earthquake monitoring technology, and in particular to a method and system for acquiring and transmitting earthquake geomagnetic observation data. Background Technology

[0002] The acquisition and transmission of seismic geomagnetic observation data is a key technological aspect of earthquake monitoring and early warning systems. Currently, this field commonly employs geomagnetic sensor arrays triggered by fixed thresholds, combined with traditional ADC sampling and wireless transmission technologies to achieve data acquisition. Conventional methods continuously acquire geomagnetic data at a fixed frequency (typically 50-200Hz) using magnetoresistive sensors or Fluxgate magnetometers. Data recording is triggered when the detected magnetic field strength exceeds a preset threshold, and the data is transmitted to a data center via LPWAN technologies such as LoRa or NB-IoT. For data processing, preprocessing schemes combining wavelet transform and FIR filtering are frequently used, along with data compression algorithms based on compressed sensing to improve transmission efficiency. These technical solutions have become standardized applications in seismic network construction, meeting basic geomagnetic anomaly monitoring needs.

[0003] However, existing methods still have room for improvement in dealing with sudden geological events. On the one hand, fixed threshold triggering mechanisms are difficult to adapt to dynamic changes in magnetic field gradients under complex environments, which may lead to missed detection of weak precursor signals or false triggering of electromagnetic interference. On the other hand, traditional data processing workflows do not sufficiently extract features from pulsed geomagnetic signals, and high-frequency components are easily attenuated during transmission, affecting the accuracy of subsequent analysis. Especially in environments with strong electromagnetic interference, the ability of existing systems to preserve the temporal characteristics of signals needs to be improved. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for acquiring and transmitting seismic geomagnetic observation data to solve the problems of accurate capture and high-fidelity transmission of geomagnetic anomalies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for acquiring and transmitting earthquake geomagnetic observation data, comprising: dynamically triggering a high-voltage pulse signal and generating an event marker data packet based on the rate of change of the environmental magnetic field gradient.

[0008] The system acquires raw waveform data with event markers, performs coordinated pulse delivery based on a spiking neural network computing environment, and obtains an enhanced pulse timing sequence.

[0009] A pulse filtering algorithm based on dynamic amplitude threshold is adopted, combined with a phase-locked loop synchronization method to optimize the information density and timing consistency of the enhanced pulse timing sequence, thereby generating an optimized pulse sequence with high signal-to-noise ratio;

[0010] The optimized pulse sequence is encapsulated into a standard transmission frame and transmitted to the earthquake detection center. At the same time, it is decoded and reconstructed using a CNN-LSTM model to generate high-fidelity earthquake and geomagnetic observation data.

[0011] As a preferred embodiment of the method for acquiring and transmitting seismic geomagnetic observation data according to the present invention, the steps of dynamically triggering a high-voltage pulse signal and generating an event-marked data packet based on the rate of change of the environmental magnetic field gradient are as follows:

[0012] The rate of change of the environmental magnetic field gradient was continuously monitored using Terfenol-D alloy sheets.

[0013] Based on statistical analysis of historical geomagnetic gradient data, a threshold for triggering effective events is defined.

[0014] When the rate of change of the ambient magnetic field gradient is greater than or equal to the effective event triggering threshold, the magnetostrictive effect triggering signal is used to drive the Terfenol-D alloy sheet to deform, and the PZT-5H piezoelectric ceramic is excited to output a high voltage pulse signal through mechanical coupling.

[0015] After the high-voltage pulse signal is excited, the UTC timestamp, instantaneous magnetic field strength change rate and background magnetic field baseline are acquired synchronously, and the data is encapsulated into an event tag data packet in TLV format through the SPI interface.

[0016] As a preferred embodiment of the method for acquiring and transmitting seismic and geomagnetic observation data according to the present invention, the acquisition of raw waveform data with event markers refers to the synchronous acquisition of raw waveform data by a triaxial seismic sensor and a triaxial geomagnetic sensor, and the real-time association of event marker data packets during the acquisition process to generate raw waveform data with event markers.

[0017] The raw waveform data with event markers includes triaxial seismic waveforms, triaxial geomagnetic waveforms, UTC timestamps, instantaneous magnetic field intensity change rate, and background magnetic field baseline.

[0018] As a preferred embodiment of the method for acquiring and transmitting seismic geomagnetic observation data according to the present invention, the steps for performing cooperative pulse delivery based on a spiking neural network computing environment to obtain an enhanced pulse time series sequence are as follows:

[0019] The Intel Loihi 2 neuromorphic chip was used to initialize a spiking neural network computing environment composed of LIF neurons;

[0020] Based on the spiking neural network computing environment, the initial waveform data with time stamps is analyzed by LIF neurons. When the dynamic event trigger threshold is exceeded, the coordinated pulse emission is triggered, generating the original pulse signal.

[0021] By using a three-level decomposition algorithm based on Daubechies4 wavelets, the firing timing and spatial distribution of the original pulse signal are analyzed at multiple scales. After extracting the time-frequency characteristic coefficients, uniform quantization coding is performed to generate an enhanced pulse timing sequence.

[0022] As a preferred embodiment of the seismic geomagnetic observation data acquisition and transmission method described in this invention, the following steps are taken: A pulse filtering algorithm based on a dynamic amplitude threshold is employed, combined with a phase-locked loop synchronization method to optimize the information density and temporal consistency of the enhanced pulse time series, generating an optimized pulse sequence with a high signal-to-noise ratio.

[0023] Based on historical geomagnetic gradient monitoring data, an amplitude threshold is defined using a sliding window statistical analysis method.

[0024] By performing a dynamic amplitude filtering algorithm on an enhanced pulse time sequence, redundant pulses below the amplitude threshold are removed, generating a pre-screened pulse sequence.

[0025] The initially screened pulse sequence is phase-calibrated using a voltage-controlled oscillator with a digital phase-locked loop, and a timing-aligned pulse sequence is output.

[0026] A pulse interval compression algorithm is applied to the time-aligned pulse sequence to increase the information density by adjusting the firing frequency, thereby generating a high-density pulse sequence.

[0027] The timing consistency of high-density pulse sequences is optimized by using a dynamic time warping algorithm, resulting in optimized pulse sequences with high signal-to-noise ratio.

[0028] As a preferred embodiment of the method for acquiring and transmitting seismic geomagnetic observation data described in this invention, the step of encapsulating the optimized pulse sequence into a standard transmission frame refers to adding a protocol header to the optimized pulse sequence and encapsulating it using the ASN.1 encoding format to generate a standard transmission frame.

[0029] As a preferred embodiment of the method for acquiring and transmitting seismic geomagnetic observation data according to the present invention, the steps for generating high-fidelity seismic geomagnetic observation data through decoding and reconstruction using a CNN-LSTM model are as follows:

[0030] Standard transmission frames are transmitted to the earthquake detection center via LPWAN. Local spatial features in the standard transmission frames are extracted by the CNN convolutional layer of the CNN-LSTM model, and long-term temporal dependencies between local spatial features are learned by the gating mechanism of the LSTM network.

[0031] Based on the long-term temporal dependencies between local spatial features, standard transmission frames are decoded and reconstructed, and the original waveform data with event markers is recovered using a pulse timing inversion algorithm.

[0032] In a second aspect, the present invention provides a seismic geomagnetic observation data acquisition and transmission system, including a data tagging module for dynamically triggering a high-voltage pulse signal and generating an event tagging data packet based on the rate of change of the environmental magnetic field gradient.

[0033] The pulse coordination enhancement module is used to acquire raw waveform data with event markers, perform coordinated pulse delivery based on the spiking neural network computing environment, and obtain enhanced pulse timing sequences.

[0034] The pulse optimization module is used to optimize the information density and timing consistency of the enhanced pulse timing sequence by using a pulse filtering algorithm based on dynamic amplitude threshold and combined with the phase-locked loop synchronization method, so as to generate an optimized pulse sequence with high signal-to-noise ratio.

[0035] The seismic waveform reconstruction module is used to encapsulate the optimized pulse sequence into a standard transmission frame and transmit it to the seismic detection center. At the same time, it uses a CNN-LSTM model to decode and reconstruct the data, generating high-fidelity seismic and geomagnetic observation data.

[0036] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the method for acquiring and transmitting seismic geomagnetic observation data as described in the first aspect of the present invention.

[0037] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for acquiring and transmitting seismic geomagnetic observation data as described in the first aspect of the present invention.

[0038] The beneficial effects of this invention are as follows: By utilizing a spiking neural network computing environment and leveraging the biological characteristics of LIF neurons, coordinated pulse firing of raw waveform data with event labels is achieved, effectively enhancing the information representation capability of pulse time-series sequences; through a CNN-LSTM model, spatial features of transmitted frames are extracted using CNN convolutional layers, and temporal dependencies are learned using an LSTM network, enabling accurate reconstruction of compressed transmitted data. These two steps work together to form a complete technology chain from data acquisition to transmission reconstruction, ensuring both real-time response capability to geomagnetic anomalies and the integrity and accuracy of observation data. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a method for acquiring and transmitting seismic and geomagnetic observation data.

[0041] Figure 2 This is a schematic diagram of a data acquisition and transmission system for earthquake and geomagnetic observations.

[0042] Figure 3 A flowchart for enhancing the synergy between event tagging and impulses. Detailed Implementation

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0044] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0046] Reference Figures 1-3 This is one embodiment of the present invention, which provides a method for acquiring and transmitting seismic geomagnetic observation data, comprising the following steps:

[0047] S1. Based on the rate of change of the environmental magnetic field gradient, dynamically trigger a high-voltage pulse signal and generate an event marker data packet;

[0048] The rate of change of the environmental magnetic field gradient was continuously monitored using a Terfenol-D alloy sheet at a sampling frequency of 100 Hz.

[0049] The change in the environmental magnetic field gradient refers to the rate of difference in magnetic field strength between two adjacent points in space (unit: nT / m·s). -1 (This is used to reflect geological activity or electromagnetic interference events;)

[0050] Based on statistical analysis of historical geomagnetic gradient data, an effective event trigger threshold is defined, typically ranging from 0.1 to 1.0 nT.

[0051] It should be noted that valid events refer to anomalies in magnetic field gradients caused by changes in rock stress due to earthquake precursors, piezomagnetic effects generated by microfractures in the Earth's crust, or local magnetic field disturbances caused by magmatic activity.

[0052] Electromagnetic interference events refer to abnormal fluctuations in the magnetic field caused by non-geological factors, such as power frequency interference generated by the operation of electrical equipment, transient magnetic field changes caused by lightning discharge, or geomagnetic disturbances caused by solar activity. These interferences can cause abnormal fluctuations in the rate of change of the environmental magnetic field gradient monitored by Terfenol-D alloy sheets, which may affect the accuracy of determining the effective event trigger threshold.

[0053] When the rate of change of the ambient magnetic field gradient is greater than or equal to the effective event trigger threshold, the magnetostrictive effect trigger signal drives the Terfenol-D alloy sheet to deform, and the PZT-5H piezoelectric ceramic is excited to output a high-voltage pulse signal through mechanical coupling.

[0054] Furthermore, when the rate of change of the environmental magnetic field gradient is detected to be greater than or equal to the effective event trigger threshold, the Terfenol-D alloy sheet generates a magnetostrictive effect under the action of the magnetic field, and the length of the alloy sheet undergoes microscopic deformation. The deformation is directly transmitted to the PZT-5H piezoelectric ceramic through the mechanical coupling structure. The piezoelectric ceramic generates polarization charge due to mechanical stress. The polarization charge is converted into a high-voltage pulse signal output by the signal amplification circuit, realizing a high-fidelity conversion from magnetic field change to electrical signal, ensuring the reliable capture and transmission of geological activity signals.

[0055] After the high-voltage pulse signal is excited, the UTC timestamp, instantaneous magnetic field strength change rate and background magnetic field baseline are acquired synchronously, and the data is encapsulated into an event tag data packet in TLV format through the SPI interface.

[0056] Furthermore, immediately after the high-voltage pulse signal is activated, a synchronous data acquisition process is initiated: a GPS receiver acquires a UTC timestamp accurate to the microsecond level, a triaxial magnetometer captures the instantaneous magnetic field strength change rate in real time, and a background magnetic field baseline is recorded as a reference. The acquired UTC timestamp, instantaneous magnetic field strength change rate, and background magnetic field baseline are transmitted to the data processing unit via the SPI interface and structured according to the TLV (Tag-Length-Value) format: the Tag field identifies the data type, the Length field records the data length, and the Value field stores the actual value. The encapsulated event tag data packet contains complete time information, magnetic field change parameters, and environmental reference data, forming a standardized data storage unit. The data packet structure ensures clear boundaries for each field, facilitating subsequent parsing and processing.

[0057] S2. Acquire raw waveform data with event markers, perform coordinated pulse delivery based on the spiking neural network computing environment, and obtain an enhanced pulse timing sequence;

[0058] Raw waveform data is acquired synchronously by a triaxial seismic sensor and a triaxial geomagnetic sensor, and event-tagged data packets are associated in real time during the acquisition process to generate raw waveform data with event tags.

[0059] Furthermore, the triaxial seismic sensor and the triaxial geomagnetic sensor synchronously start data acquisition at a sampling frequency of 100Hz. The triaxial seismic sensor records the seismic waveforms along the X / Y / Z axes, while the triaxial geomagnetic sensor synchronously acquires the magnetic field waveforms along the X / Y / Z axes. During the acquisition process, event-marked data packets triggered by the Terfenol-D alloy sheet are received in real time. The seismic waveform data, geomagnetic waveform data, and the UTC timestamp, instantaneous magnetic field intensity change rate, and background magnetic field baseline in the event-marked data packets are time-aligned and parameter-correlated to generate raw waveform data with event marks, including triaxial seismic waveforms, triaxial geomagnetic waveforms, UTC timestamps, instantaneous magnetic field intensity change rates, and background magnetic field baselines.

[0060] The raw waveform data with event markers includes triaxial seismic waveforms, triaxial geomagnetic waveforms, UTC timestamps, instantaneous magnetic field intensity change rate, and background magnetic field baseline;

[0061] The Intel Loihi 2 neuromorphic chip was used to initialize a spiking neural network computing environment consisting of 128 LIF neurons;

[0062] Furthermore, the specific implementation process of initializing the 128 LIF neuromorphic computing environment on the Intel Loihi 2 neuromorphic chip is as follows: When configuring the neuronal network through the NxCore instruction set, the chip first establishes a two-dimensional grid topology, enabling physically adjacent neurons to form local connection clusters. Based on this, fundamental parameters such as resting potential, firing threshold, and membrane time constant are set for each LIF neuron to ensure that the neuron possesses biologically reasonable dynamic response characteristics. Subsequently, the STDP rule is used to initialize synaptic connection weights, where excitatory synapses promote signal propagation, and inhibitory synapses maintain network balance. The asynchronous spiking event routing engine optimizes the spiking transmission path based on a priority queue, working in conjunction with neuron parameter configuration to jointly construct a low-latency communication network. The x86 coprocessor monitors the network status in real time through a dedicated bus, dynamically adjusting the global clock frequency based on neuron firing frequency and energy consumption data, forming a closed-loop control mechanism. From physical connections and parameter configuration to dynamic control, the above process ultimately constructs a biologically plastic spiking neural network computing environment, achieving efficient spatiotemporal pattern processing capabilities.

[0063] It should be noted that this is a plasticity learning mechanism based on pulse timing dependence. For example, when the presynaptic neuron pulse arrives earlier than the postsynaptic neuron pulse, the synaptic weight is enhanced; conversely, the weight is weakened.

[0064] Based on the spiking neural network computing environment, the time-stamped initial waveform data is analyzed through 28 LIF neurons. When the dynamic event trigger threshold is exceeded, the coordinated pulse emission is triggered, generating the original pulse signal.

[0065] Furthermore, the Intel Loihi 2 neuromorphic chip receives time-stamped initial waveform data input, and 28 LIF neurons synchronously process the data according to a preset two-dimensional grid topology. Each LIF neuron continuously integrates the input pulses, and immediately fires a pulse when the membrane potential exceeds the dynamic event triggering threshold. The fired pulses propagate to neighboring neurons through synaptic connections adjusted by STDP rules, triggering a network cascade response. The neuron cluster that meets the coordinated firing condition synchronously outputs the original pulse signal through an asynchronous pulse event routing engine, with the signal timing accuracy strictly synchronized with the time stamp of the original waveform data. This process realizes the pulse encoding conversion of the initial waveform data, providing a precise spatiotemporal feature representation basis for the subsequent generation of enhanced pulse time series sequences.

[0066] It should be noted that the dynamic event triggering threshold is defined based on statistical analysis of historical geomagnetic gradient data, and the effective event triggering threshold range is usually 0.1-1.0 nT.

[0067] Co-firing condition refers to the cluster response mechanism triggered when multiple LIF neurons in a spiking neural network successively reach the firing threshold within a specific time window (usually 5-10ms). For example, when analyzing earthquake precursor signals, if more than 60% of the 28 LIF neurons fire synchronously within an 8ms time window, it is considered a valid co-firing event.

[0068] By using a three-level decomposition algorithm based on Daubechies4 wavelets, the firing timing and spatial distribution of the original pulse signal are analyzed at multiple scales. After extracting the time-frequency characteristic coefficients, uniform quantization coding is performed to generate an enhanced pulse timing sequence.

[0069] Furthermore, the emission timing of the original pulse signal is first subjected to a three-level wavelet decomposition. The first level of decomposition extracts features in the 0-50Hz frequency band, the second level focuses on the 50-100Hz frequency band, and the third level analyzes the 100-200Hz frequency band. Each level of decomposition simultaneously extracts the spatial distribution features of the pulse signal along the X / Y / Z axes, generating time-frequency feature coefficients containing both time-frequency and spatial information (this enables multi-scale feature representation of the original pulse signal in the time-frequency domain). After normalization, the time-frequency feature coefficients are discretized into integer values ​​of 0-255 using 8-bit uniform quantization encoding. The quantized feature coefficients are then sorted and reassembled according to timestamps, ultimately outputting an enhanced pulse timing sequence with multi-scale time-frequency features. This sequence fully preserves the spatiotemporal pattern features of the original pulse signal while enhancing the resolvability of high-frequency components.

[0070] It should be noted that the expression for extracting the time-frequency characteristic coefficients is as follows:

[0071]

[0072] Where C is the time-frequency characteristic coefficient, N is the total number of effective wavelet coefficients, j is the decomposition scale level, k is the translation factor, S is the original pulse signal, and ψ j,k This represents the wavelet component of the Daubechies4 wavelet basis function at the j-th decomposition scale and the k-th translation position.

[0073] Effective wavelet coefficients refer to the detail coefficients that are not contaminated by boundary effects after Daubechies4 wavelet decomposition. Their number is determined by the signal length and the decomposition level.

[0074] It should be noted that the original pulse signal S is standardized by Z-score before being used in the calculation to eliminate sensor gain differences and ensure that the time-frequency characteristic coefficient C is dimensionless and numerically stable.

[0075] S3. A pulse filtering algorithm based on dynamic amplitude threshold is adopted, combined with a phase-locked loop synchronization method to optimize the information density and timing consistency of the enhanced pulse timing sequence, and generate an optimized pulse sequence with high signal-to-noise ratio.

[0076] Based on historical geomagnetic gradient monitoring data, an amplitude threshold is defined using a sliding window statistical analysis method.

[0077] Furthermore, time windows are divided into fixed durations; within each window, the mean and standard deviation of the magnetic field gradient change rate are calculated using the sliding window statistical analysis method; and an amplitude threshold is established by analyzing the statistical characteristics of window data from different geological activity periods, typically ranging from 0.1 to 1.0 nT.

[0078] By performing a dynamic amplitude filtering algorithm on an enhanced pulse time sequence, redundant pulses below the amplitude threshold are removed, generating a pre-screened pulse sequence.

[0079] Furthermore, the enhanced pulse time series generated by Daubechies4 wavelet decomposition is read first, and the amplitude value and release timestamp of each pulse event are analyzed. The pulse amplitude value is compared in real time with the dynamic amplitude threshold defined based on historical geomagnetic gradient monitoring data to establish pulse validity judgment conditions. A sliding window detection mechanism is used to scan all pulses within a 10ms time window and remove redundant pulses with amplitude values ​​lower than the dynamic amplitude threshold. The retained pulse events are reordered according to the original timestamps to generate a preliminarily filtered pulse sequence.

[0080] The initially screened pulse sequence is phase-calibrated using a voltage-controlled oscillator with a digital phase-locked loop to output a timing-aligned pulse sequence.

[0081] Furthermore, the digital phase-locked loop (PLL) first detects the phase deviation of the input pulse sequence, and the voltage-controlled oscillator (VCO) dynamically adjusts the oscillation frequency based on the error signal output by the phase detector. When a pulse timing offset is detected, the VCO precisely adjusts the phase of the output pulse by changing the control voltage, with each adjustment step being 0.1 ns. After 5-10 cycles of continuous phase tracking and correction, the timing jitter of the pulse sequence is controlled within ±2 ns. The final output timing-aligned pulse sequence maintains the amplitude characteristics and waveform features of the original pulses, with only the emission time being fine-tuned to ensure that the rising edge of all pulses is strictly aligned with the reference clock edge. During phase calibration, the bandwidth parameter of the PLL is adaptively adjusted according to the input pulse rate, with the bandwidth range set to 100 kHz-1 MHz to balance tracking speed and stability.

[0082] A pulse interval compression algorithm is applied to the time-aligned pulse sequence to increase the information density by adjusting the firing frequency, thereby generating a high-density pulse sequence.

[0083] Furthermore, the distribution characteristics of the original firing intervals of the time-aligned pulse sequence are first analyzed, and a pulse interval histogram is established. Based on the histogram statistical results, the maximum allowable compression ratio is determined. While maintaining the integrity of the pulse waveform, a nonlinear time scaling algorithm is used to gradually reduce the pulse interval. During the compression process, the pulse amplitude distortion is dynamically monitored, and the current compression operation is automatically stopped when the distortion exceeds 5%. The compressed pulse sequence is then re-calibrated through a digital phase-locked loop to eliminate the timing jitter introduced by time scaling. The final high-density pulse sequence maintains the amplitude characteristics and waveform integrity of the original pulses. The compression algorithm is implemented using a sliding window, with the window size adaptively adjusted according to the pulse rate, ranging from 10 to 50 pulse cycles.

[0084] The timing consistency of high-density pulse sequences is optimized by using a dynamic time warping algorithm, resulting in optimized pulse sequences with high signal-to-noise ratio.

[0085] Furthermore, firstly, a temporal correspondence between the high-density pulse sequence and the reference template sequence is established. The dynamic time warping algorithm calculates the minimum cumulative path distance between each pulse event and the template sequence. The algorithm adjusts the pulse firing time through nonlinear time warping to eliminate the time distortion introduced by pulse interval compression. During optimization, a dynamic programming method is used to solve for the optimal time warping path, with constraints including a maximum time offset not exceeding 50 ns and a path slope maintained within the range of 0.5-2.0. The timing-adjusted pulse sequence undergoes final phase calibration through a digital phase-locked loop to ensure that the rising edge of all pulses is strictly synchronized with the system clock. The final optimized pulse sequence maintains the amplitude characteristics and waveform integrity of the original pulses. The window length of the dynamic time warping algorithm is adaptively adjusted according to the pulse density, ranging from 10 to 30 pulse cycles.

[0086] Template sequences refer to typical pulse pattern reference sequences generated by clustering historical data. They contain standardized timestamps and amplitude features and are used for time alignment calculations in dynamic time warping.

[0087] The minimum cumulative path distance between each impulse event and the template sequence is calculated using the following expression:

[0088]

[0089] Where, d n This represents the minimum cumulative alignment distance between the first n events of the input pulse sequence and the template sequence, where α is a time-weighted coefficient (typically ranging from 10 to 10). 4 ~10 8 ), β is the amplitude-dimension weighting coefficient (usually ranging from 0.1 to 1.0), t m Here, is the timestamp of the m-th event in the input sequence, M is the total number of events in the template sequence, and l is the index of the current event in the template sequence. It is the template event timestamp aligned to the m-th input event on path π, a m It is the normalized amplitude of the m-th event in the input pulse sequence. It is the template event amplitude aligned to the m-th input event on path π. It is the set of all valid aligned paths from the starting point (1,1) to the current point (n,l).

[0090] It should be noted that t m and Before being used in the calculation, all data were normalized to eliminate the difference between time and amplitude dimensions and ensure the stability of the distance calculation values.

[0091] S4. The optimized pulse sequence is encapsulated into a standard transmission frame and transmitted to the earthquake detection center. At the same time, it is decoded and reconstructed using a CNN-LSTM model to generate high-fidelity earthquake and geomagnetic observation data.

[0092] A protocol header is added to the optimized pulse sequence, and it is encapsulated using the ASN.1 encoding format to generate a standard transmission frame;

[0093] Furthermore, a 16-byte protocol header is added before the optimized pulse sequence. This header includes a 4-byte synchronization word, a 4-byte sequence length identifier, a 4-byte timestamp, and a 4-byte checksum. The protocol header and optimized pulse sequence are structured and encapsulated using the ASN.1 Basic Encoding Rule (BER), with each field encoded according to the TLV (Type-Length-Value) triple format. During encapsulation, each sampling point of the pulse sequence is encoded using the ASN.1 REAL type, and the timestamp is encoded using the ASN.1 GeneralizedTime type. The encoded data unit generates a 4-byte frame check sequence using the CRC-32 algorithm, which is appended to the ASN.1 encoded data to form a complete standard transmission frame. The total length of the standard transmission frame dynamically varies according to the number of sampling points in the optimized pulse sequence, with a minimum frame length of 64 bytes and a maximum of 1024 bytes.

[0094] Standard transmission frames are transmitted to the earthquake detection center via LPWAN. Local spatial features in the standard transmission frames are extracted using the CNN convolutional layer of the CNN-LSTM model, and long-term temporal dependencies between local spatial features are learned through the gating mechanism of the LSTM network.

[0095] Furthermore, the CNN-LSTM model employs a three-layer convolutional kernel structure in its CNN convolutional layers to process pulse sequences. The first layer uses 16 5×1 convolutional kernels to extract local spatial features, the second layer uses 32 3×1 convolutional kernels to capture higher-level spatial patterns, and the third layer uses 64 3×1 convolutional kernels to refine features. Each convolutional layer is followed by a ReLU activation function and max pooling, ultimately outputting a 128-dimensional spatial feature vector. The LSTM network's gating mechanism processes the feature sequence output by the CNN: the input gate controls feature updates, the forget gate manages long-term state memory, and the output gate adjusts the intensity of feature expression. LSTM units establish long-term dependencies between local spatial features through recurrent connections.

[0096] Local spatial features include pulse waveform temporal gradient features, triaxial sensor spatial coupling features, and multi-scale frequency domain energy features.

[0097] It should be noted that the CNN-LSTM model has synergistic advantages over traditional single-modal neural networks. Specifically, the CNN convolutional layer extracts the temporal gradient features and three-axis spatial coupling features of the pulse sequence through 5×1 / 3×1 multi-scale convolutional kernels, while the LSTM network gating mechanism establishes long-term dependencies over a time span of 50 sampling points. The two work together to achieve joint optimization of spatiotemporal features, and the final output contains feature representations that simultaneously include waveform details and evolutionary patterns.

[0098] Based on the long-term temporal dependencies between local spatial features, standard transmission frames are decoded and reconstructed, and the original waveform data with event markers is recovered using a pulse timing inversion algorithm.

[0099] Furthermore, the spatiotemporal feature vector is decomposed into local spatial feature components and long-term temporal dependent feature components through principal component analysis. The local spatial feature components are used to reconstruct the spatial distribution patterns of the triaxial seismic sensor and the triaxial geomagnetic sensor through deconvolution operation, while the long-term temporal dependent feature components are used to reconstruct the pulse emission time series marked with UTC timestamps through backpropagation of the LSTM network. The pulse time series inversion algorithm fuses the reconstructed spatial distribution patterns with the temporal information, and through iterative optimization, matches the time-frequency characteristics of the reconstructed pulse sequence with the original pulse sequence, ultimately restoring and generating the original waveform data marked with events.

[0100] This embodiment also provides a seismic geomagnetic observation data acquisition and transmission system, including: a data tagging module, used to dynamically trigger high-voltage pulse signals and generate event tagging data packets according to the rate of change of the environmental magnetic field gradient;

[0101] The pulse coordination enhancement module is used to acquire raw waveform data with event tags, perform coordinated pulse delivery based on the spiking neural network computing environment, and obtain enhanced pulse timing sequences.

[0102] The pulse optimization module is used to optimize the information density and timing consistency of the enhanced pulse timing sequence by using a pulse filtering algorithm based on dynamic amplitude threshold and combined with the phase-locked loop synchronization method, so as to generate an optimized pulse sequence with high signal-to-noise ratio.

[0103] The seismic waveform reconstruction module is used to encapsulate the optimized pulse sequence into a standard transmission frame and transmit it to the seismic detection center. At the same time, it uses a CNN-LSTM model to decode and reconstruct the data, generating high-fidelity seismic and geomagnetic observation data.

[0104] This embodiment also provides a computer device applicable to the method of acquiring and transmitting seismic and geomagnetic observation data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the method for acquiring and transmitting seismic and geomagnetic observation data as proposed in the above embodiment.

[0105] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0106] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the method for acquiring and transmitting seismic geomagnetic observation data as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0107] In summary, this invention utilizes a spiking neural network computing environment, leveraging the biological characteristics of LIF neurons to achieve coordinated pulse firing of event-tagged raw waveform data, effectively enhancing the information representation capability of pulse time-series sequences. Furthermore, through a CNN-LSTM model, spatial features of transmitted frames are extracted using CNN convolutional layers, and temporal dependencies are learned using an LSTM network, enabling accurate reconstruction of compressed transmitted data. These two steps work together to form a complete technology chain from data acquisition to transmission reconstruction, ensuring both real-time response capabilities to geomagnetic anomalies and the integrity and accuracy of observational data.

[0108] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for acquiring and transmitting seismic geomagnetic observation data, characterized in that: include, Based on the rate of change of the environmental magnetic field gradient, a high-voltage pulse signal is dynamically triggered and an event marker data packet is generated. The process involves acquiring raw waveform data with event markers, performing coordinated pulse firing based on a spiking neural network computing environment, and obtaining an enhanced pulse timing sequence. The steps are as follows: The Intel Loihi 2 neuromorphic chip was used to initialize a spiking neural network computing environment composed of LIF neurons; Based on the spiking neural network computing environment, the initial waveform data with time stamps is analyzed by LIF neurons. When the dynamic event trigger threshold is exceeded, the coordinated pulse emission is triggered, generating the original pulse signal. By using a three-level decomposition algorithm based on Daubechies4 wavelets, the firing timing and spatial distribution of the original pulse signal are analyzed at multiple scales. After extracting the time-frequency feature coefficients, uniform quantization coding is performed to generate an enhanced pulse timing sequence. A pulse filtering algorithm based on dynamic amplitude thresholding, combined with a phase-locked loop synchronization method, is used to optimize the information density and temporal consistency of the enhanced pulse timing sequence, generating an optimized pulse sequence with high signal-to-noise ratio. The steps are as follows: Based on historical geomagnetic gradient monitoring data, an amplitude threshold is defined using a sliding window statistical analysis method. By performing a dynamic amplitude filtering algorithm on an enhanced pulse time sequence, redundant pulses below the amplitude threshold are removed, generating a pre-screened pulse sequence. The initially screened pulse sequence is phase-calibrated using a voltage-controlled oscillator with a digital phase-locked loop, and a timing-aligned pulse sequence is output. A pulse interval compression algorithm is applied to the time-aligned pulse sequence to increase the information density by adjusting the firing frequency, thereby generating a high-density pulse sequence. The timing consistency of high-density pulse sequences is optimized by using a dynamic time warping algorithm to generate optimized pulse sequences with high signal-to-noise ratio. The optimized pulse sequence is encapsulated into a standard transmission frame and transmitted to the earthquake detection center. At the same time, it is decoded and reconstructed using a CNN-LSTM model to generate high-fidelity earthquake and geomagnetic observation data.

2. The method for acquiring and transmitting seismic geomagnetic observation data as described in claim 1, characterized in that: The steps for dynamically triggering a high-voltage pulse signal and generating an event marker data packet based on the rate of change of the environmental magnetic field gradient are as follows: The rate of change of the environmental magnetic field gradient was continuously monitored using Terfenol-D alloy sheets. Based on statistical analysis of historical geomagnetic gradient data, a threshold for triggering effective events is defined. When the rate of change of the ambient magnetic field gradient is greater than or equal to the effective event triggering threshold, the magnetostrictive effect triggering signal is used to drive the Terfenol-D alloy sheet to deform, and the PZT-5H piezoelectric ceramic is excited to output a high voltage pulse signal through mechanical coupling. After the high-voltage pulse signal is excited, the UTC timestamp, instantaneous magnetic field strength change rate and background magnetic field baseline are acquired synchronously, and the data is encapsulated into an event tag data packet in TLV format through the SPI interface.

3. The method for acquiring and transmitting seismic geomagnetic observation data as described in claim 1, characterized in that: Acquiring raw waveform data with event tags refers to synchronously acquiring raw waveform data using a triaxial seismic sensor and a triaxial geomagnetic sensor, and associating event tag data packets in real time during the acquisition process to generate raw waveform data with event tags. The raw waveform data with event markers includes triaxial seismic waveforms, triaxial geomagnetic waveforms, UTC timestamps, instantaneous magnetic field intensity change rate, and background magnetic field baseline.

4. The method for acquiring and transmitting seismic geomagnetic observation data as described in claim 1, characterized in that: The process of encapsulating the optimized pulse sequence into a standard transmission frame refers to adding a protocol header to the optimized pulse sequence and encapsulating it using the ASN.1 encoding format to generate a standard transmission frame.

5. The method for acquiring and transmitting seismic geomagnetic observation data as described in claim 4, characterized in that: The steps for decoding and reconstructing high-fidelity seismic and geomagnetic observation data using a CNN-LSTM model are as follows: Standard transmission frames are transmitted to the earthquake detection center via LPWAN. Local spatial features in the standard transmission frames are extracted by the CNN convolutional layer of the CNN-LSTM model, and long-term temporal dependencies between local spatial features are learned by the gating mechanism of the LSTM network. Based on the long-term temporal dependencies between local spatial features, standard transmission frames are decoded and reconstructed, and the original waveform data with event markers is recovered using a pulse timing inversion algorithm.

6. A seismic geomagnetic observation data acquisition and transmission system, based on the seismic geomagnetic observation data acquisition and transmission method according to any one of claims 1 to 5, characterized in that: include, The data tagging module is used to dynamically trigger high-voltage pulse signals and generate event tagging data packets based on the rate of change of the environmental magnetic field gradient. The pulse coordination enhancement module is used to acquire raw waveform data with event markers, perform coordinated pulse delivery based on the spiking neural network computing environment, and obtain enhanced pulse timing sequences. The pulse optimization module is used to optimize the information density and timing consistency of the enhanced pulse timing sequence by using a pulse filtering algorithm based on dynamic amplitude threshold and combined with the phase-locked loop synchronization method, so as to generate an optimized pulse sequence with high signal-to-noise ratio. The seismic waveform reconstruction module is used to encapsulate the optimized pulse sequence into a standard transmission frame and transmit it to the seismic detection center. At the same time, it uses a CNN-LSTM model to decode and reconstruct the data, generating high-fidelity seismic and geomagnetic observation data.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for acquiring and transmitting seismic geomagnetic observation data as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for acquiring and transmitting seismic geomagnetic observation data as described in any one of claims 1 to 5.

Images