A distributed optical fiber detection method and system based on Brillouin and Rayleigh scattering analysis algorithm

Abstract

This invention relates to the field of fiber optic detection technology, and discloses a distributed fiber optic detection method and system based on Brillouin and Rayleigh scattering analysis algorithms. The method includes splitting a probe beam and a local oscillator beam from the same narrow-linewidth laser beam, and multiplexing and transmitting them according to measurement frames. This allows the same sensing fiber to obtain Rayleigh scattering coherent echo signals and Brillouin scattering signals. The Rayleigh signal is demodulated to obtain vibration information and event location. The Brillouin signal is frequency-shifted and inverted to obtain a temperature field or static strain field. A mapping model between temperature, static strain, and Rayleigh phase drift is established. Drift compensation and baseline calibration are performed on the vibration signal. Distance calibration, time synchronization, anti-aliasing filtering, interpolation resampling, and fusion identification are performed on the two-channel data. The event type, location, and corresponding temperature or strain change are output, along with graded alarms and visualization. This invention achieves dual-scattering collaborative detection, improving identification accuracy and early warning capabilities.

Description

Technical Field

[0001] This invention relates to the field of optical fiber detection technology, and in particular to a distributed optical fiber detection method and system based on Brillouin and Rayleigh scattering analysis algorithms. Background Technology

[0002] Distributed fiber optic sensing technology is a type of sensing technology that utilizes the scattering characteristics of light signals in optical fibers to continuously detect external disturbances, temperature changes, and strain changes at various locations along the fiber. Compared with traditional point sensors, distributed fiber optic sensing technology has advantages such as long sensing distance, large monitoring range, strong anti-electromagnetic interference capability, and applicability to complex environments. Therefore, it is widely used in fields such as oil and gas pipeline monitoring, perimeter security, power cable monitoring, rail transit monitoring, bridge and tunnel health monitoring, and geological disaster early warning. In existing distributed fiber optic sensing technologies, Rayleigh scattering-based detection methods are typically used to acquire vibration or dynamic strain information distributed along the fiber. These technologies generally achieve the location and identification of external disturbances by coherently demodulating and analyzing the backscattered echoes. Because Rayleigh scattering signals are sensitive to weak vibrations, they have high application value in scenarios such as construction disturbance detection, intrusion alarms, and vehicle passing identification. However, in practical applications, Rayleigh scattering-based distributed fiber optic sensing technology is easily affected by factors such as laser phase noise, ambient temperature changes, fiber static strain changes, transmit power drift, local oscillator phase drift, and receiver gain drift. This leads to phase baseline drift, decreased detection sensitivity, and increased false alarm rate. Especially in long-distance, long-term continuous monitoring scenarios, the drift caused by these slow variables significantly affects the stable extraction of dynamic vibration signals, thus impacting event detection accuracy and identification reliability. On the other hand, Brillouin scattering-based distributed fiber optic sensing technology is typically used to measure temperature and static strain information distributed along the fiber. Existing Brillouin detection methods typically involve acquiring the Brillouin gain spectrum and extracting the Brillouin frequency shift parameters, then converting these parameters into corresponding temperatures or strains. This type of technology can reflect the slowly changing state of the measured object and is significant in scenarios such as structural health monitoring and temperature early warning. However, existing Brillouin scattering detection techniques generally suffer from slow measurement speeds, complex frequency shift extraction processes, and strong dependence on signal-to-noise ratio and spectral fitting accuracy. When the detection environment is complex, the scattering spectrum is distorted, or there is significant noise, the stability and accuracy of the Brillouin frequency shift inversion results are easily affected, thus impacting the measurement results of the temperature and static strain fields. Furthermore, in existing technologies, Rayleigh scattering detection and Brillouin scattering detection are mostly implemented as independent technical solutions. They are often separate in terms of transmission methods, demodulation methods, time references, distance coordinates, and data processing flows, making effective collaboration difficult under the same sensing fiber, time reference, and distance coordinates. This makes it difficult to use the temperature and static strain information obtained from the Brillouin channel to compensate for the phase drift in the Rayleigh channel, and also makes it difficult to perform unified correlation analysis of vibration, temperature and strain information, thus limiting the comprehensive detection capability in complex scenarios. Therefore, there is an urgent need for a distributed optical fiber detection method and system based on Brillouin and Rayleigh scattering analysis algorithms. Summary of the Invention

[0003] The purpose of this invention is to overcome one or more of the above-mentioned existing technical problems and provide a distributed optical fiber detection method and system based on Brillouin and Rayleigh scattering analysis algorithms.

[0004] To achieve the above objectives, this invention provides a distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms, comprising: The probe light and the local oscillator light are obtained by splitting the same narrow linewidth laser beam. The probe light is encoded and multiplexed and transmitted according to the measurement frame, so that the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Coherent demodulation of Rayleigh scattering coherent echo signals yields complex echo sequences along the distance axis. At least one of dynamic strain time series and vibration time series is obtained by differential phase, phase expansion and band-limited filtering of adjacent sampling points. Event center and spatial location are determined based on energy mutation, cross-correlation peak or sparse matching tracking. A wideband response of the Brillouin scattering signal is obtained at a fixed radio frequency center frequency. A random phase sequence is generated and an inverse Fourier transform of the spectrum is performed to obtain a wideband Brillouin time-domain response. The initial value of the Brillouin frequency shift is obtained based on a neural network. The corrected frequency shift is obtained by physical constraint fitting through Lorentz or Gaussian spectral type. At least one of the temperature field and static strain field is converted. A mapping model of temperature and static strain on Rayleigh phase drift is established, and drift compensation and baseline calibration of vibration signals are performed based on adaptive filtering, recursive least squares or Kalman estimation. Distance calibration, time synchronization, anti-aliasing filtering, and interpolation resampling are performed on the two-channel data. A preset time window is extracted according to the event center to construct a three-dimensional tensor of space, time, and channel. After standardization and normalization, the tensor is input into the fusion recognition model of convolutional network and temporal network to obtain the event type, location, and at least one of the corresponding temperature change and strain change. The system then performs graded alarms and visualization based on the confidence level.

[0005] According to one aspect of the present invention, one or more of pulse coding, phase coding or linear frequency modulation coding are employed, and the code length and pulse width are set to improve processing gain and meet the target spatial resolution. Rayleigh and Brillouin detector sequences are transmitted alternately in time-division multiplexing within the same measurement frame, or the two types of detector sequences are loaded with orthogonal polarizations in polarization multiplexing and polarization diversity demultiplexed at the receiving end. A reference segment is set in the optical fiber as the system self-calibration benchmark. The echo amplitude and phase stability of the reference segment is used to periodically correct the drift of the transmit power, the drift of the local oscillator phase, and the drift of the receive gain, so that the Rayleigh channel and the Brillouin channel can be compared and fused under the same distance coordinates and the same timestamp.

[0006] According to one aspect of the present invention, a sliding window is established according to the spatial position of the complex echo sequence and the differential phase is calculated. After phase unwrapping, bandpass filtering is performed to separate the target frequency band. The time spectrum is obtained by performing short-time Fourier transform, wavelet packet decomposition or empirical mode decomposition on the filtered vibration signal, and the vibration feature vector is formed by calculating features such as multi-band energy, spectral centroid, peak frequency, bandwidth, kurtosis, spectral entropy, cepstral coefficient and autocorrelation period. Environmental noise and sudden interference are suppressed by adaptive thresholding, power frequency notch filtering, spectral subtraction, sparse representation or robust PCA. Fixed-duration samples are extracted from the event center by pre-set sampling points before and after the event center to make the input dimension of different events consistent. The fixed-duration samples are 50 sampling points before and after the event center and correspond to a time length of 60 seconds. The samples are also z-score normalized and 0-1 normalized.

[0007] According to one aspect of the present invention, the time-domain curve corresponding to a single broadband Brillouin gain spectrum is acquired under the condition of a fixed radio frequency center frequency, and the time-frequency domain scattering spectrum is multiplied by a random phase sequence and then subjected to inverse Fourier transform to obtain a Brillouin time-domain response signal with broadband characteristics. The neural network is a convolutional neural network or a lightweight Transformer. The network input is the multi-scale features of the time-domain response, and the output is the initial value of the Brillouin frequency shift and its uncertainty. Physical constraint fitting at least constrains the range of values ​​for spectral linewidth, center frequency, and noise floor, and employs coarse-to-fine iteration and confidence testing. When the fitting residual exceeds the limit, historical data is called for verification or resampling is performed. The neural network dataset construction involves applying at least one of known temperature and strain and simultaneously adjusting device parameters to establish a correlation between frequency shift and scattering spectrum. Sample selection is based on a genetic algorithm to cover different signal-to-noise ratio conditions.

[0008] According to one aspect of the present invention, at least one of the temperature field and the static strain field obtained by Brillouin frequency shift inversion is used as a slow variable input, a Rayleigh phase drift prediction term is established based on the fiber phase temperature coefficient and strain coefficient, and the drift model is updated on the time axis using recursive least squares or Kalman filtering. When the Brillouin frequency shift is affected by both temperature and strain, temperature and strain are decoupled based on the cross-sensitivity coefficient, and then the decoupled slow variable is used for drift compensation. A reference fiber segment is set as the zero-point calibration, and the phase baseline and gain drift are periodically corrected. The compensated vibration signal is then subjected to weak event enhancement and false alarm suppression. Weak event enhancement includes spectral whitening, matched filtering, or wavelet thresholding for denoising. False alarm suppression includes statistical testing of the compensation residual and joint judgment with at least one consistency constraint among slow variables of temperature and slow variables of strain.

[0009] According to one aspect of the invention, the three-dimensional tensor includes at least two of the vibration channel, temperature channel and strain channel, and introduces adaptive weights for different channels based on signal-to-noise ratio, fitting quality factor, uncertainty or historical stability. The fusion recognition model includes a convolutional layer for extracting spatial correlation, an LSTM layer for extracting temporal dependence, and an attention layer for cross-channel fusion. The fusion recognition model outputs at least one of the following: event category, event start and end time, spatial location, temperature change, and strain change, along with its confidence level. Based on the confidence level and a preset threshold, it performs graded alarms, false alarm rollback, and event tracking display. The uncertainty of Brillouin demodulation and the signal-to-noise ratio of vibration features are input into the attention layer for probability fusion, and the event probability distribution, Top-k candidate categories and corresponding handling suggestions are given at the output.

[0010] According to one aspect of the present invention, the training of the fusion recognition model includes establishing an event sample library and extracting sampling points before and after the event center according to uniform rules to form a fixed-length sample, and performing data augmentation on the sample by normalization, noise superposition, time shift, amplitude scaling, frequency band perturbation, and random occlusion. During training, a multi-task joint optimization of classification loss and location regression loss is adopted, and hard sample mining, semi-supervised pseudo-labels or domain adaptive constraints are introduced to reduce the distribution shift caused by different laying methods, different fiber attenuation and different background noise. High-confidence new samples are stored in the sample library and incrementally updated to adapt the fusion recognition model to the field environment. The event types include at least one or more of the following: mechanical construction, manual construction, rainwater erosion, vehicle passage, and subway passage. After training, the hierarchical alarm thresholds are determined through cross-validation, and false alarm feedback and missed alarm feedback are recorded on the deployment end for proactive learning and updating.

[0011] To achieve the above objectives, this invention provides a distributed optical fiber detection system based on Brillouin and Rayleigh scattering analysis algorithms, comprising: Signal acquisition module: The probe light and local oscillator light are obtained by splitting the same narrow linewidth laser beam. The probe light is encoded and multiplexed and transmitted according to the measurement frame, so that the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Rayleigh scattering coherent echo signal processing module: performs coherent demodulation on Rayleigh scattering coherent echo signal to obtain complex echo sequence along the distance axis, obtains at least one of dynamic strain time series and vibration time series through differential phase, phase expansion and band-limited filtering of adjacent sampling points, and determines the event center and spatial location based on energy mutation, cross-correlation peak or sparse matching tracking. Brillouin scattering signal processing module: Obtains wideband response of Brillouin scattering signal at fixed radio frequency center frequency, generates random phase sequence and performs inverse Fourier transform of spectrum to obtain wideband Brillouin time domain response, obtains initial value of Brillouin frequency shift based on neural network, obtains corrected frequency shift by physical constraint fitting through Lorentz or Gaussian spectrum, and converts at least one of temperature field and static strain field. Mapping model building module: Establishes a mapping model of temperature and static strain on Rayleigh phase drift, and performs drift compensation and baseline calibration on vibration signals based on adaptive filtering, recursive least squares or Kalman estimation; Information warning and visualization module: Perform distance calibration, time synchronization, anti-aliasing filtering and interpolation resampling on the two-channel data, and extract a preset time window according to the event center to construct a three-dimensional tensor of space, time and channel. After standardization and normalization, it is input into the fusion recognition model of convolutional network and temporal network to obtain the event type, location and at least one of the corresponding temperature change and strain change, and perform hierarchical alarm and visualization according to confidence level.

[0012] To achieve the above objectives, the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the above-described distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms.

[0013] To achieve the above objectives, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms.

[0014] Based on this, the beneficial effects of the present invention are as follows: by splitting the probe light and the local oscillator light from the same narrow linewidth laser beam, and by using the measurement frame multiplexing method, the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Therefore, it is convenient to realize the correspondence, comparison and fusion of the two types of scattering signals under the same distance coordinate and the same time stamp, which is beneficial to improving the consistency and synergy of distributed fiber multi-parameter detection. By establishing a Rayleigh phase drift mapping model using at least one of the temperature field and static strain field obtained by Brillouin frequency shift inversion, and combining adaptive filtering, recursive least squares or Kalman estimation to perform drift compensation and baseline calibration on the vibration signal, it is beneficial to reduce the influence of environmental temperature changes, static strain changes and system slow drift on the Rayleigh vibration detection results, and improve the stability and reliability of the detection results. By performing distance calibration, time synchronization, and resampling on Rayleigh and Brillouin channel data, a three-dimensional tensor of space, time, and channel is constructed. This tensor is then input into a fusion recognition model to achieve a joint output of event type, location, and at least one of temperature and strain changes. Furthermore, hierarchical alarms and visualizations are performed based on confidence levels, which helps improve the accuracy and practicality of event recognition and early warning in complex scenarios. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating a distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms according to an exemplary embodiment; Figure 2 This is a flowchart illustrating a distributed optical fiber detection system based on Brillouin and Rayleigh scattering analysis algorithms according to an exemplary embodiment. Detailed Implementation

[0016] The invention will now be discussed with reference to exemplary embodiments. It should be understood that the described embodiments are merely intended to enable those skilled in the art to better understand and thus implement the invention, and are not intended to imply any limitation on the scope of the invention.

[0017] As used herein, the term “comprising” and its variations are to be interpreted as open-ended terms meaning “including but not limited to”. The term “based on” is to be interpreted as “at least partially based on”, and the terms “one embodiment” and “an embodiment” are to be interpreted as “at least one embodiment”.

[0018] According to one embodiment of the present invention, Figure 1 This is a flowchart illustrating a distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms according to an exemplary embodiment, such as... Figure 1 As shown, to achieve the above objectives, this invention provides a distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms, comprising: The probe light and the local oscillator light are obtained by splitting the same narrow linewidth laser beam. The probe light is encoded and multiplexed and transmitted according to the measurement frame, so that the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Coherent demodulation of Rayleigh scattering coherent echo signals yields complex echo sequences along the distance axis. At least one of dynamic strain time series and vibration time series is obtained by differential phase, phase expansion and band-limited filtering of adjacent sampling points. Event center and spatial location are determined based on energy mutation, cross-correlation peak or sparse matching tracking. A wideband response of the Brillouin scattering signal is obtained at a fixed radio frequency center frequency. A random phase sequence is generated and an inverse Fourier transform of the spectrum is performed to obtain a wideband Brillouin time-domain response. The initial value of the Brillouin frequency shift is obtained based on a neural network. The corrected frequency shift is obtained by physical constraint fitting through Lorentz or Gaussian spectral type. At least one of the temperature field and static strain field is converted. A mapping model of temperature and static strain on Rayleigh phase drift is established, and drift compensation and baseline calibration of vibration signals are performed based on adaptive filtering, recursive least squares or Kalman estimation. Distance calibration, time synchronization, anti-aliasing filtering, and interpolation resampling are performed on the two-channel data. A preset time window is extracted according to the event center to construct a three-dimensional tensor of space, time, and channel. After standardization and normalization, the tensor is input into the fusion recognition model of convolutional network and temporal network to obtain the event type, location, and at least one of the corresponding temperature change and strain change. The system then performs graded alarms and visualization based on the confidence level.

[0019] According to one embodiment of the present invention, one or more of pulse coding, phase coding or linear frequency modulation coding are used to set the code length and pulse width to improve processing gain and meet the target spatial resolution; Rayleigh and Brillouin detector sequences are transmitted alternately in time-division multiplexing within the same measurement frame, or the two types of detector sequences are loaded with orthogonal polarizations in polarization multiplexing and polarization diversity demultiplexed at the receiving end. A reference segment is set in the optical fiber as the system self-calibration benchmark. The echo amplitude and phase stability of the reference segment is used to periodically correct the drift of the transmit power, the drift of the local oscillator phase, and the drift of the receive gain, so that the Rayleigh channel and the Brillouin channel can be compared and fused under the same distance coordinates and the same timestamp.

[0020] According to one embodiment of the present invention, a sliding window is established for the complex echo sequence according to its spatial position and the differential phase is calculated. After phase unwrapping, bandpass filtering is performed to separate the target frequency band. The time spectrum is obtained by performing short-time Fourier transform, wavelet packet decomposition or empirical mode decomposition on the filtered vibration signal, and the vibration feature vector is formed by calculating features such as multi-band energy, spectral centroid, peak frequency, bandwidth, kurtosis, spectral entropy, cepstral coefficient and autocorrelation period. Environmental noise and sudden interference are suppressed by adaptive thresholding, power frequency notch filtering, spectral subtraction, sparse representation or robust PCA. Fixed-duration samples are extracted from the event center by pre-set sampling points before and after the event center to make the input dimension of different events consistent. The fixed-duration samples are 50 sampling points before and after the event center and correspond to a time length of 60 seconds. The samples are also z-score normalized and 0-1 normalized.

[0021] According to one embodiment of the present invention, the time-domain curve corresponding to a single broadband Brillouin gain spectrum is acquired under a fixed radio frequency center frequency condition, and the time-frequency domain scattering spectrum is multiplied by a random phase sequence and then subjected to inverse Fourier transform to obtain a Brillouin time-domain response signal with broadband characteristics. The neural network is a convolutional neural network or a lightweight Transformer. The network input is the multi-scale features of the time-domain response, and the output is the initial value of the Brillouin frequency shift and its uncertainty. Physical constraint fitting at least constrains the range of values ​​for spectral linewidth, center frequency, and noise floor, and employs coarse-to-fine iteration and confidence testing. When the fitting residual exceeds the limit, historical data is called for verification or resampling is performed. The neural network dataset construction involves applying at least one of known temperature and strain and simultaneously adjusting device parameters to establish a correlation between frequency shift and scattering spectrum. Sample selection is based on a genetic algorithm to cover different signal-to-noise ratio conditions.

[0022] According to one embodiment of the present invention, at least one of the temperature field and static strain field obtained by Brillouin frequency shift inversion is used as a slow variable input, a Rayleigh phase drift prediction term is established based on the fiber phase temperature coefficient and strain coefficient, and the drift model is updated on the time axis using recursive least squares or Kalman filtering. When the Brillouin frequency shift is affected by both temperature and strain, temperature and strain are decoupled based on the cross-sensitivity coefficient, and then the decoupled slow variable is used for drift compensation. A reference fiber segment is set as the zero-point calibration, and the phase baseline and gain drift are periodically corrected. The compensated vibration signal is then subjected to weak event enhancement and false alarm suppression. Weak event enhancement includes spectral whitening, matched filtering, or wavelet thresholding for denoising. False alarm suppression includes statistical testing of the compensation residual and joint judgment with at least one consistency constraint among slow variables of temperature and slow variables of strain.

[0023] According to one embodiment of the present invention, the three-dimensional tensor includes at least two of the vibration channel, temperature channel and strain channel, and an adaptive weight based on signal-to-noise ratio, fitting quality factor, uncertainty or historical stability is introduced for different channels. The fusion recognition model includes a convolutional layer for extracting spatial correlation, an LSTM layer for extracting temporal dependence, and an attention layer for cross-channel fusion. The fusion recognition model outputs at least one of the following: event category, event start and end time, spatial location, temperature change, and strain change, along with its confidence level. Based on the confidence level and a preset threshold, it performs graded alarms, false alarm rollback, and event tracking display. The uncertainty of Brillouin demodulation and the signal-to-noise ratio of vibration features are input into the attention layer for probability fusion, and the event probability distribution, Top-k candidate categories and corresponding handling suggestions are given at the output.

[0024] According to one embodiment of the present invention, the training of the fusion recognition model includes establishing an event sample library and extracting sampling points before and after the event center according to a unified rule to form a fixed-length sample, and performing data augmentation on the sample by normalization, noise superposition, time shift, amplitude scaling, frequency band perturbation, and random occlusion. During training, a multi-task joint optimization of classification loss and location regression loss is adopted, and hard sample mining, semi-supervised pseudo-labels or domain adaptive constraints are introduced to reduce the distribution shift caused by different laying methods, different fiber attenuation and different background noise. High-confidence new samples are stored in the sample library and incrementally updated to adapt the fusion recognition model to the field environment. The event types include at least one or more of the following: mechanical construction, manual construction, rainwater erosion, vehicle passage, and subway passage. After training, the hierarchical alarm thresholds are determined through cross-validation, and false alarm feedback and missed alarm feedback are recorded on the deployment end for proactive learning and updating.

[0025] Specifically, in the Rayleigh channel, the Rayleigh scattering coherent echo signal is coherently demodulated to obtain a complex echo sequence distributed along the distance axis. Then, through differential phase, phase expansion, and band-limited filtering of adjacent sampling points, dynamic strain time series and / or vibration time series are obtained. Then, the event center and spatial location are determined based on energy mutation, cross-correlation peak, or sparse matched tracking. To adapt to the noise environment on site, adaptive thresholding, power frequency notch filtering, spectral subtraction, sparse representation, or robust PCA can be combined to suppress environmental noise and sudden interference. For the detected event segment, a fixed time window is truncated based on the event center, for example, 50 sampling points before and after the event center are truncated and corresponding to a time length of 60 seconds, so that different event samples have a unified input dimension.

[0026] Specifically, in the Brillouin channel, a wideband response is acquired at a fixed RF center frequency. A random phase sequence is generated and subjected to an inverse Fourier transform of the spectrum to obtain a wideband Brillouin time-domain response signal. Then, a neural network is used to analyze the multi-scale characteristics of this time-domain response, outputting the initial value of the Brillouin frequency shift and its uncertainty. This initial value is then used as the initial parameter for physical constraint fitting, and a Lorentz or Gaussian spectral model is used for fitting to obtain the corrected Brillouin frequency shift. Based on the corrected Brillouin frequency shift, the temperature field and static strain field distributed along the optical fiber can be further calculated. The combination of neural network initial estimation and physical model fitting correction reduces the sensitivity of traditional fitting to initial values ​​and maintains high demodulation accuracy under noisy conditions.

[0027] Specifically, in the two-channel fusion stage, the temperature field and static strain field obtained by Brillouin frequency shift inversion are used as slow variable inputs. A Rayleigh phase drift prediction term is established based on the fiber phase temperature coefficient and strain coefficient. Adaptive filtering, recursive least squares, or Kalman filtering are used to update the drift model to achieve drift compensation and baseline calibration of the Rayleigh vibration signal. When the Brillouin frequency shift is simultaneously affected by temperature and strain coupling, temperature and strain can be decoupled using the cross-sensitivity coefficient. The decoupled result is then used for compensation. Through this compensation mechanism, the interference of factors such as sudden temperature changes, slow load changes, and light source drift on vibration event identification can be reduced.

[0028] Specifically, the data obtained from the Rayleigh and Brillouin channels undergo distance calibration, time synchronization, anti-aliasing filtering, and interpolation resampling. A preset time window is then extracted based on the event center to construct a three-dimensional tensor containing spatial, temporal, and channel dimensions. This three-dimensional tensor can include at least two of the vibration, temperature, and strain channels, and adaptive weights can be assigned based on the signal-to-noise ratio, fitting quality factor, uncertainty, or historical stability of each channel. The three-dimensional tensor is then standardized and normalized before being input into the fusion recognition model. The fusion recognition model can include convolutional layers for extracting spatial correlations, LSTM layers for extracting temporal dependencies, and attention layers for cross-channel fusion. The final output includes event category, spatial location, event start and end times, temperature change, strain change, and their confidence levels. Graded alarms, false alarm rollback, and event tracking display are then implemented according to preset thresholds.

[0029] Specifically, during the training phase, an event sample library tailored to engineering scenarios can be established. The event types in the sample library should include at least one or more of the following: mechanical construction, manual construction, rainwater erosion, vehicle passage, and subway passage. Simultaneously, multi-dimensional information such as vibration, stress, and temperature can be stored. To enhance the model's adaptability to different laying methods, background noise levels, and site conditions, data augmentation techniques such as noise superposition, time shifting, amplitude scaling, frequency band perturbation, and random occlusion can be employed during training. Furthermore, classification loss and location regression loss can be used for multi-task joint optimization. After deployment, new high-confidence samples can be continuously added to the sample library for incremental updates and active learning, thereby improving the system's long-term adaptability to the site environment.

[0030] Specifically, sensing optical fibers are laid along urban rail transit lines, around stations, or near tunnels to perform distributed detection of disturbance events around the lines. Since subway passage and mechanical construction can both generate significant vibration responses, relying solely on a single vibration signal can easily lead to misjudgments. Therefore, a joint identification method using simultaneous detection of Rayleigh scattering and Brillouin scattering channels is employed. The system first acquires Rayleigh scattering coherent echo signals and Brillouin scattering signals within the same measurement frame, and completes time synchronization and distance alignment. For the Rayleigh channel, the vibration timing and corresponding time-frequency characteristics caused by train operation or external construction are extracted; for the Brillouin channel, the temperature field and static strain field along the line are acquired, and Rayleigh phase drift is compensated accordingly. Subsequently, a fixed-length sample is extracted based on the event center to construct a three-dimensional tensor containing vibration, temperature, and strain, which is then input into the fusion recognition model. In this scenario, the passing of the subway typically exhibits vibration characteristics that propagate continuously along the line direction, while mechanical construction typically exhibits strong and relatively concentrated impact vibrations within a local section. If the local temperature changes and static strain changes reflected by the Brillouin Channel are combined simultaneously, the distinguishability between the two can be further improved. Thus, the system can output the category, location, and corresponding confidence level of the subway passing event or mechanical construction event, and issue graded alarms accordingly.

[0031] Specifically, sensing optical fibers are laid along integrated utility tunnels, oil and gas pipelines, or other underground linear infrastructure to identify events such as vehicle passage, manual construction, and mechanical construction. Since surface traffic loads, manual knocking, and mechanical excavation can all cause soil vibration near pipelines, and environmental temperature changes and slow-varying stress can cause Rayleigh channel baseline drift, the temperature field and static strain field information provided by the Brillouin channel are used to compensate and correct the Rayleigh vibration results. The complex echo sequence collected by the Rayleigh channel is demodulated, filtered, and feature extracted to obtain the vibration feature vector of the event to be identified; the Brillouin channel obtains the Brillouin frequency shift of the corresponding section by combining neural network initial estimation with physical constraint fitting, and calculates the temperature field and static strain field. Subsequently, a Rayleigh phase drift prediction term is constructed based on the temperature and strain coefficients, and the drift model is updated using recursive least squares or Kalman filtering to output the compensated vibration signal. In this scenario, passing vehicles typically exhibit short-duration passing vibrations, mechanical construction typically exhibits strong, continuous impact vibrations, and manual construction exhibits low-amplitude but rhythmic local vibrations. By inputting the compensated vibration signal along with temperature and strain information into the fusion recognition model, the model can distinguish between passing vehicles, mechanical construction, and manual construction, thereby reducing the false alarm rate.

[0032] Specifically, sensing optical fibers are laid along riverbanks, slopes, mountain passes, or shallow buried lines to detect events such as rainwater erosion, vehicle passage, and on-site emergency repairs. In this scenario, the continuous low-frequency disturbances caused by rainwater erosion and the vibrations generated by passing vehicles are similar under certain conditions. Therefore, multi-physics joint analysis is used to improve event discrimination capabilities. Vibration information is obtained through the Rayleigh channel, and temperature and static strain fields are obtained through the Brillouin channel. Then, the data from the two channels are synchronized in time, calibrated at a distance, and subjected to a unified window to form a three-dimensional tensor input fusion recognition model. Since rainwater erosion generates continuous vibrations, it may also be accompanied by local temperature changes and slow strain changes, while vehicle passage usually exhibits short-term passing vibrations with relatively weak static strain changes. Therefore, combining Brillouin channel information can effectively improve the separability of the two. When the confidence level of the rainwater erosion event output by the model exceeds the preset threshold, an alarm can be triggered in conjunction with the event location, temperature change, and strain change. When the model outputs vehicle passage or construction events, different alarm levels or handling suggestions are adopted according to the corresponding categories. Through the above methods, the long-term online monitoring needs in complex natural environments can be met.

[0033] Furthermore, to achieve the aforementioned objectives, this invention also provides a distributed optical fiber detection system based on Brillouin and Rayleigh scattering analysis algorithms. Figure 2 This is a flowchart illustrating a distributed optical fiber detection system based on Brillouin and Rayleigh scattering analysis algorithms according to an exemplary embodiment, such as... Figure 2 As shown, a distributed optical fiber detection system based on Brillouin and Rayleigh scattering analysis algorithms in this invention includes: Signal acquisition module: The probe light and local oscillator light are obtained by splitting the same narrow linewidth laser beam. The probe light is encoded and multiplexed and transmitted according to the measurement frame, so that the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Rayleigh scattering coherent echo signal processing module: performs coherent demodulation on Rayleigh scattering coherent echo signal to obtain complex echo sequence along the distance axis, obtains at least one of dynamic strain time series and vibration time series through differential phase, phase expansion and band-limited filtering of adjacent sampling points, and determines the event center and spatial location based on energy mutation, cross-correlation peak or sparse matching tracking. Brillouin scattering signal processing module: Obtains wideband response of Brillouin scattering signal at fixed radio frequency center frequency, generates random phase sequence and performs inverse Fourier transform of spectrum to obtain wideband Brillouin time domain response, obtains initial value of Brillouin frequency shift based on neural network, obtains corrected frequency shift by physical constraint fitting through Lorentz or Gaussian spectrum, and converts at least one of temperature field and static strain field. Mapping model building module: Establishes a mapping model of temperature and static strain on Rayleigh phase drift, and performs drift compensation and baseline calibration on vibration signals based on adaptive filtering, recursive least squares or Kalman estimation; Information warning and visualization module: Perform distance calibration, time synchronization, anti-aliasing filtering and interpolation resampling on the two-channel data, and extract a preset time window according to the event center to construct a three-dimensional tensor of space, time and channel. After standardization and normalization, it is input into the fusion recognition model of convolutional network and temporal network to obtain the event type, location and at least one of the corresponding temperature change and strain change, and perform hierarchical alarm and visualization according to confidence level.

[0034] To achieve the above-mentioned objectives, the present invention also provides an electronic device comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the above-mentioned distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms.

[0035] To achieve the above-mentioned objectives, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms.

[0036] Those skilled in the art will recognize that the modules and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0037] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and equipment can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0038] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0039] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0040] In addition, the functional modules in the embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0041] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the energy-saving signal transmission / reception methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0042] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

[0043] It should be understood that the sequence number of each step in the invention and embodiments of the present invention does not absolutely imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

Claims

1. A distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms, characterized in that, include: The probe light and the local oscillator light are obtained by splitting the same narrow linewidth laser beam. The probe light is encoded and multiplexed and transmitted according to the measurement frame, so that the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Coherent demodulation of Rayleigh scattering coherent echo signals yields complex echo sequences along the distance axis. At least one of dynamic strain time series and vibration time series is obtained by differential phase, phase expansion and band-limited filtering of adjacent sampling points. Event center and spatial location are determined based on energy mutation, cross-correlation peak or sparse matching tracking. A wideband response of the Brillouin scattering signal is obtained at a fixed radio frequency center frequency. A random phase sequence is generated and an inverse Fourier transform of the spectrum is performed to obtain a wideband Brillouin time-domain response. The initial value of the Brillouin frequency shift is obtained based on a neural network. The corrected frequency shift is obtained by physical constraint fitting through Lorentz or Gaussian spectral type. At least one of the temperature field and static strain field is converted. A mapping model of temperature and static strain on Rayleigh phase drift is established, and drift compensation and baseline calibration of vibration signals are performed based on adaptive filtering, recursive least squares or Kalman estimation. Distance calibration, time synchronization, anti-aliasing filtering, and interpolation resampling are performed on the two-channel data. A preset time window is extracted according to the event center to construct a three-dimensional tensor of space, time, and channel. After standardization and normalization, the tensor is input into the fusion recognition model of convolutional network and temporal network to obtain the event type, location, and at least one of the corresponding temperature change and strain change. The system then performs graded alarms and visualization based on the confidence level.

2. The distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in claim 1, characterized in that, One or more of pulse coding, phase coding, or linear frequency modulation coding are used, and the code length and pulse width are set to improve processing gain and meet the target spatial resolution. Rayleigh and Brillouin detector sequences are transmitted alternately in time-division multiplexing within the same measurement frame, or the two types of detector sequences are loaded with orthogonal polarizations in polarization multiplexing and polarization diversity demultiplexed at the receiving end. A reference segment is set in the optical fiber as the system self-calibration benchmark. The echo amplitude and phase stability of the reference segment is used to periodically correct the drift of the transmit power, the drift of the local oscillator phase, and the drift of the receive gain, so that the Rayleigh channel and the Brillouin channel can be compared and fused under the same distance coordinates and the same timestamp.

3. The distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in claim 2, characterized in that, A sliding window is established for the complex echo sequence according to its spatial location and the differential phase is calculated. After phase unwrapping, bandpass filtering is performed to separate the target frequency band. The time spectrum is obtained by performing short-time Fourier transform, wavelet packet decomposition or empirical mode decomposition on the filtered vibration signal, and the vibration feature vector is formed by calculating features such as multi-band energy, spectral centroid, peak frequency, bandwidth, kurtosis, spectral entropy, cepstral coefficient and autocorrelation period. Environmental noise and sudden interference are suppressed by adaptive thresholding, power frequency notch filtering, spectral subtraction, sparse representation or robust PCA. Fixed-duration samples are extracted from the event center by pre-set sampling points before and after the event center to make the input dimension of different events consistent. The fixed-duration samples are 50 sampling points before and after the event center and correspond to a time length of 60 seconds. The samples are also z-score normalized and 0-1 normalized.

4. The distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in claim 3, characterized in that, Under a fixed radio frequency center frequency, the time-domain curve corresponding to a single broadband Brillouin gain spectrum is acquired, and the time-frequency domain scattering spectrum is multiplied by a random phase sequence and then subjected to inverse Fourier transform to obtain a Brillouin time-domain response signal with broadband characteristics. The neural network is a convolutional neural network or a lightweight Transformer. The network input is the multi-scale features of the time-domain response, and the output is the initial value of the Brillouin frequency shift and its uncertainty. Physical constraint fitting at least constrains the range of values ​​for spectral linewidth, center frequency, and noise floor, and employs coarse-to-fine iteration and confidence testing. When the fitting residual exceeds the limit, historical data is called for verification or resampling is performed. The neural network dataset construction involves applying at least one of known temperature and strain and simultaneously adjusting device parameters to establish a correlation between frequency shift and scattering spectrum. Sample selection is based on a genetic algorithm to cover different signal-to-noise ratio conditions.

5. The distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in claim 4, characterized in that, Using at least one of the temperature field and static strain field obtained by Brillouin frequency shift inversion as slow variable input, a Rayleigh phase drift prediction term is established based on the fiber phase temperature coefficient and strain coefficient, and the drift model is updated on the time axis using recursive least squares or Kalman filtering. When the Brillouin frequency shift is affected by both temperature and strain, temperature and strain are decoupled based on the cross-sensitivity coefficient, and then the decoupled slow variable is used for drift compensation. A reference fiber segment is set as the zero-point calibration, and the phase baseline and gain drift are periodically corrected. The compensated vibration signal is then subjected to weak event enhancement and false alarm suppression. Weak event enhancement includes spectral whitening, matched filtering, or wavelet thresholding for denoising. False alarm suppression includes statistical testing of the compensation residual and joint judgment with at least one consistency constraint among slow variables of temperature and slow variables of strain.

6. The distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in claim 5, characterized in that, The three-dimensional tensor contains at least two of the vibration, temperature and strain channels, and introduces adaptive weights based on signal-to-noise ratio, fitting quality factor, uncertainty or historical stability for different channels. The fusion recognition model includes a convolutional layer for extracting spatial correlation, an LSTM layer for extracting temporal dependence, and an attention layer for cross-channel fusion. The fusion recognition model outputs at least one of the following: event category, event start and end time, spatial location, temperature change, and strain change, along with its confidence level. Based on the confidence level and a preset threshold, it performs graded alarms, false alarm rollback, and event tracking display. The uncertainty of Brillouin demodulation and the signal-to-noise ratio of vibration features are input into the attention layer for probability fusion, and the event probability distribution, Top-k candidate categories and corresponding handling suggestions are given at the output.

7. The distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in claim 6, characterized in that, The training of the fusion recognition model includes establishing an event sample library and extracting sampling points before and after the event center according to a unified rule to form a fixed-length sample. The sample is then normalized, noise superimposed, time-shifted, amplitude scaled, frequency band perturbed, and randomly occluded for data augmentation. During training, a multi-task joint optimization of classification loss and location regression loss is adopted, and hard sample mining, semi-supervised pseudo-labels or domain adaptive constraints are introduced to reduce the distribution shift caused by different laying methods, different fiber attenuation and different background noise. High-confidence new samples are stored in the sample library and incrementally updated to adapt the fusion recognition model to the field environment. The event types include at least one or more of the following: mechanical construction, manual construction, rainwater erosion, vehicle passage, and subway passage. After training, the hierarchical alarm thresholds are determined through cross-validation, and false alarm feedback and missed alarm feedback are recorded on the deployment end for proactive learning and updating.

8. A distributed optical fiber detection system based on Brillouin and Rayleigh scattering analysis algorithms, characterized in that, include: Signal acquisition module: The probe light and local oscillator light are obtained by splitting the same narrow linewidth laser beam. The probe light is encoded and multiplexed and transmitted according to the measurement frame, so that the same sensing fiber can obtain Rayleigh scattering coherent echo signal and Brillouin scattering signal under the same time reference. Rayleigh scattering coherent echo signal processing module: performs coherent demodulation on Rayleigh scattering coherent echo signal to obtain complex echo sequence along the distance axis, obtains at least one of dynamic strain time series and vibration time series through differential phase, phase expansion and band-limited filtering of adjacent sampling points, and determines the event center and spatial location based on energy mutation, cross-correlation peak or sparse matching tracking. Brillouin scattering signal processing module: Obtains wideband response of Brillouin scattering signal at fixed radio frequency center frequency, generates random phase sequence and performs inverse Fourier transform of spectrum to obtain wideband Brillouin time domain response, obtains initial value of Brillouin frequency shift based on neural network, obtains corrected frequency shift by physical constraint fitting through Lorentz or Gaussian spectrum, and converts at least one of temperature field and static strain field. Mapping model building module: Establishes a mapping model of temperature and static strain on Rayleigh phase drift, and performs drift compensation and baseline calibration on vibration signals based on adaptive filtering, recursive least squares or Kalman estimation; Information warning and visualization module: Perform distance calibration, time synchronization, anti-aliasing filtering and interpolation resampling on the two-channel data, and extract a preset time window according to the event center to construct a three-dimensional tensor of space, time and channel. After standardization and normalization, it is input into the fusion recognition model of convolutional network and temporal network to obtain the event type, location and at least one of the corresponding temperature change and strain change, and perform hierarchical alarm and visualization according to confidence level.

9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements a distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements a distributed optical fiber detection method based on Brillouin and Rayleigh scattering analysis algorithms as described in any one of claims 1 to 7.

Images