A distributed optical fiber vibration signal demodulation and positioning method based on intelligent optimization algorithm

Abstract

This invention belongs to the field of distributed optical fiber sensing and signal processing technology, and relates to a method for demodulating and locating optical fiber vibration signals suitable for long-distance pipeline intrusion monitoring. Addressing the problems of high false alarm rate, insufficient positioning accuracy, and mutual constraints between frequency response and sensing distance in existing phase-sensitive optical time-domain reflectometers (φ-OTDRs) under conditions of large data volumes and low signal-to-noise ratios, this invention combines algorithms such as I / Q demodulation, amplitude differential positioning, phase dewinding, and coherent fading suppression from φ-OTDRs with an intelligent parameter optimization method based on the bat algorithm to achieve adaptive optimization of demodulation parameters. The system employs a narrow-linewidth laser and a backscattering Rayleigh scattering coherent detection structure. I / Q demodulation data acquisition is completed via a PCIE6920 acquisition card. On the software side, amplitude demodulation algorithms including direct differential, continuous averaging and differential, and moving average and moving differential are used for coarse localization. Phase demodulation algorithms such as range expansion, initial phase subtraction, spatial dewinding, temporal dewinding, and downsampling differential dewinding are used to recover the vibration waveform. Coherent fading suppression and adaptive filtering are performed under intelligently optimized parameter combinations, achieving a spatial resolution of 10 m and a vibration frequency response of 0–10 kHz within a 30 km pipeline range. Compared to traditional fixed-parameter demodulation methods, this invention can automatically adjust demodulation parameters under different environments and pipeline operating conditions, improving positioning accuracy and frequency response range, and reducing false alarm and false negative rates. It is suitable for intrusion monitoring of long-distance oil and gas pipelines and other important linear infrastructure.

Description

Technical Field

[0001] This invention belongs to the field of fiber optic sensing and signal processing technology, specifically relating to a distributed fiber optic vibration signal demodulation and positioning method based on intelligent optimization algorithms, which can be applied to intrusion monitoring and status monitoring of linear targets such as oil and gas pipelines, long-distance power cables, and boundary fences. Background Technology

[0002] A phase-sensitive optical time domain reflectometer (φ-OTDR) is a sensing technology that utilizes backscattered Rayleigh signals in optical fibers to achieve distributed vibration monitoring. Compared with traditional electrical sensors, distributed optical fiber sensors have advantages such as resistance to electromagnetic interference, corrosion resistance, low loss, large detection range, and the ability to achieve continuous distributed measurements, making them particularly suitable for scenarios such as long-distance pipeline intrusion monitoring. In a typical φ-OTDR system, a narrow-linewidth laser is modulated into an optical pulse by a modulation unit and coupled into a sensing fiber. A minute vibration at any point in the fiber will alter the phase and amplitude of the backscattered Rayleigh light in that vicinity. By using a coherent detection structure and a high-speed data acquisition unit, the backscattered signal corresponding to the spatial location can be obtained, thereby enabling vibration localization and waveform reconstruction. In existing technologies, φ-OTDR data processing mainly includes steps such as I / Q demodulation, amplitude differential positioning, phase dewinding, and coherent fading suppression. The vibration location can be determined using amplitude demodulation algorithms such as direct differential, continuous averaging and differential, and moving average and moving differential. The vibration waveform can be recovered using phase demodulation algorithms such as range expansion, initial phase subtraction, spatial dewinding, temporal dewinding, differential dewinding, downsampling dewinding, and downsampling differential dewinding. Simultaneously, coherent fading can be suppressed using methods such as multi-band scattered signal synthesis or vector rotation. However, in engineering applications, existing distributed fiber optic vibration monitoring methods still have the following problems: (1) Parameters are dependent on experience and have poor adaptability. The amplitude differential window length, filtering parameters, downsampling factor, threshold value, etc. are usually manually set by engineers based on their experience. When the pipe material, burial method, environmental noise or fiber length changes, the original parameter combination is often no longer applicable, resulting in increased positioning error or false alarm rate. (2) It is difficult to balance long distance and high resolution and wide bandwidth. To achieve spatial resolution of 10 m, a narrow optical pulse width is required, while long-distance sensing requires sufficient optical energy; in order to cover the frequency range of 0 to 10 kHz, the pulse repetition frequency must be high, but the high repetition frequency is limited by the length of the optical fiber, and there is a constraint between the two. (3) Huge amount of data and insufficient real-time performance. At a monitoring distance of 30 km, in order to meet the requirements of spatial resolution and frequency response, the system needs a high sampling rate and a large number of sampling points. The amount of data in a single scan is very large. The fixed algorithm is easy to cause excessive delay when processed on the CPU, which is not conducive to real-time alarm. (4) Significant coherent fading affects phase demodulation. The vector superposition of Rayleigh scattering signals will produce "fading points" with amplitude close to zero at some locations. At these locations, it is easy to obtain incorrect results when calculating the phase using the arctangent function. Without effective fading detection and correction, it will seriously affect the recovery of vibration waveforms and event recognition. Therefore, there is a need for a distributed optical fiber vibration signal demodulation and localization method that can unify the modeling of the φ-OTDR demodulation process, automatically search for the optimal signal processing parameters, and take into account long distance, high resolution, wide bandwidth and real-time performance. Summary of the Invention

[0003] (I) Purpose of the Invention The purpose of this invention is to overcome the shortcomings of existing distributed optical fiber vibration monitoring systems, such as reliance on manual experience for demodulation parameters, difficulty in balancing a 30 km sensing distance, approximately 10 m spatial resolution, and a 0–10 kHz frequency response range, as well as demodulation errors caused by coherent fading and difficulties in real-time processing. This invention proposes a distributed optical fiber vibration signal demodulation and localization method based on an intelligent optimization algorithm. By uniformly modeling the key parameters in the processing flow, including I / Q demodulation, amplitude differential demodulation, phase dewinding, and coherent fading suppression, and employing the bat algorithm for global optimization, the localization accuracy and waveform reconstruction quality of vibration events are improved while ensuring real-time performance.

[0004] (II) Technical Solution The core of the technical solution of this invention is to combine various signal processing algorithms of phase-sensitive optical time-domain reflectometer with the intelligent parameter optimization of the bat algorithm to achieve adaptive demodulation and precise positioning of distributed optical fiber vibration signals. The overall process is divided into eight key steps, and each step works together to complete the entire process from signal acquisition to real-time monitoring: 1. Detection of light pulse emission and reception of backscattered Rayleigh signals The laser output from a narrow linewidth continuous laser is modulated into a detection light pulse with a preset pulse width and repetition frequency by an acousto-optic or electro-optic modulator. The pulse propagates along the sensing optical fiber laid on the monitored pipeline, generating backscattered Rayleigh light in the process. Subsequently, a photodetector converts this optical signal into an electrical signal, completing the initial conversion from optical signal to electrical signal. 2. I / Q Demodulation and Data Acquisition The electrical signal output from the photodetector is beat-frequencyd with the local oscillator light using a coherent detection structure to obtain an intermediate frequency (IF) signal. The IF signal is then sampled using a PCIE6920 acquisition card or a high-speed acquisition card with equivalent performance. The card's built-in digital down-conversion and I / Q demodulation modules are then invoked to obtain I-channel and Q-channel data distributed along the fiber optic distance, providing basic data support for subsequent amplitude and phase calculations. 3. Amplitude demodulation and coarse vibration localization Based on the collected I / Q data, an amplitude sequence is calculated. At least one of the following methods is used to differentially demodulate the amplitude sequence: direct difference, continuous averaging and difference, or moving average and moving difference, amplifying the amplitude variation characteristics caused by vibration. By calculating the absolute value of the amplitude variation and comparing it with a preset vibration detection threshold, locations where the amplitude variation exceeds the threshold are selected, forming a set of suspected vibration locations. This completes the coarse localization of the vibration event, narrowing down the area for subsequent precise localization. 4. Phase unwinding and vibration waveform recovery The phase sequence is calculated based on I / Q data. The phase sequence is then subjected to bidirectional spatial and temporal dewinding to eliminate 2π periodic phase jumps in both spatial and temporal dimensions, restoring a continuous phase change trend. For high-frequency vibrations or conditions with drastic phase changes, the phase sequence is first downsampled. Dewinding is then performed on the downsampled sequence, and the result is mapped back to the original sampling points using interpolation, reducing computational complexity while maintaining demodulation accuracy. Finally, within the set of suspected vibration locations, the dewinded phase sequence is digitally filtered and denoised to extract the vibration displacement or strain waveform at the target location, thus restoring the vibration waveform. 5. Coherent Fading Detection and Correction Using a set distance window as a scale, the minimum and mean values ​​of the amplitude sequence are calculated within each window. A preset dual-threshold criterion is used to identify amplitude troughs caused by coherent fading. For the marked coherent fading intervals, a continuous phase curve is constructed using phase points at non-fading locations on both sides of the interval through polynomial fitting or spline interpolation. The phase values ​​corresponding to this curve are then used to replace the erroneous phase values ​​within the fading interval, effectively eliminating spurious phase interference caused by coherent fading and ensuring the accuracy of phase demodulation. 6. Unified modeling of demodulation parameters and construction of objective function Key parameters in each step of amplitude demodulation, phase dewinding, and coherent fading suppression, such as amplitude demodulation window length, spatial moving average window length, phase downsampling factor, filter parameters, and various decision thresholds, are integrated to construct a unified parameter vector. Calibration data containing multiple typical vibration events are selected to calculate three performance indicators: average positioning error, vibration waveform correlation coefficient, and signal-to-noise ratio improvement. The processing time for a single frame of data is also statistically analyzed. After normalizing all indicators, they are linearly combined according to preset weights to construct an objective function with the parameter vector as the variable, thus setting a clear optimization objective for subsequent parameter optimization. 7. Parameter optimization based on the bat algorithm The bat algorithm is employed to globally optimize the constructed parameter vector. First, hyperparameters such as the number of individual bats, maximum number of iterations, frequency range, initial loudness, and emission probability are initialized. Initial parameter vector positions and velocities for each bat are randomly generated. In each iteration, the frequency, velocity, and position of each bat are updated according to rules, and the parameter vector is constrained to a physically meaningful range. Based on the updated parameter vector, a complete demodulation process is executed, the objective function value is calculated, and the globally optimal parameter vector is updated. A local perturbation mechanism is introduced to perform a local search near the globally optimal solution, improving optimization accuracy. Based on the comparison between the random number and the individual's loudness and emission probability, it is determined whether to accept a new solution, and the loudness and emission probability of each bat are updated simultaneously. This process is repeated until the maximum number of iterations is reached or the objective function converges, outputting the globally optimal demodulation parameter vector. 8. Real-time monitoring and online adjustment The optimal parameter vector obtained by the bat algorithm is applied to the actual monitoring process. I / Q data is acquired in real time via a high-speed acquisition card. Following the demodulation and positioning process described above, the distributed fiber optic vibration signal is processed in real time. During long-term system operation, new monitoring data is periodically acquired, or when a significant change in the statistical characteristics of environmental noise is detected, the parameter optimization process is automatically re-executed to achieve online adaptive adjustment of demodulation parameters, ensuring that the system maintains good performance under different operating conditions and environments.

[0005] (III) Beneficial Effects Compared with existing technologies, this invention achieves performance improvement and functional optimization of distributed fiber optic vibration monitoring through algorithm fusion and intelligent optimization, and has the following five core beneficial effects: 1. Intelligent automatic parameter adjustment significantly improves system robustness. By integrating key parameters from each signal processing stage into an optimizable vector and employing the Bat Algorithm for global optimization, the system completely eliminates reliance on human experience. It can automatically match the optimal parameter combination based on actual fiber optic conditions, environmental noise, pipeline operating conditions, and other factors, significantly improving adaptability and robustness in different application scenarios and effectively reducing positioning errors and false alarm rates caused by parameter mismatches. 2. Balancing multiple performance indicators and overcoming technological limitations. Under the intelligent optimization framework, the repetition frequency of the probe light pulse, the number of sampling points, and the phase downsampling strategy are configured in a coordinated manner. Without adding a complex optical amplification structure, an effective monitoring distance of 30km, a spatial resolution better than 10m, and a wide frequency response of 0 to 10kHz can be achieved simultaneously. This successfully overcomes the mutual constraints between long distance, high resolution, and wide bandwidth in existing technologies. 3. Effectively suppresses coherent fading and improves demodulation and positioning accuracy. By using dual-threshold determination of amplitude statistical features, coherent fading intervals can be accurately identified. Combined with interpolation fitting, the phase of the fading interval is corrected, effectively eliminating the influence of pseudo-phase on phase dewinding and waveform recovery. This makes the phase change characteristics of the vibration position clearer, significantly reducing the false alarm rate and false alarm rate of the system, and improving the overall demodulation and positioning accuracy. 4. Optimize the computation process to ensure real-time processing performance. By optimizing algorithms such as downsampling dewinding and differential dewinding, the computational complexity is significantly reduced while ensuring demodulation accuracy. Combined with the hardware architecture of FPGA+CPU collaborative processing platform or industrial computer, the target of single-frame data processing response time of no more than 0.5s is achieved, meeting the needs of real-time alarm and rapid response in engineering applications. 5. Simple implementation method, easy to implement and integrate into projects. The hardware required for this invention consists of conventional devices in the field of fiber optic sensing, including narrow-linewidth lasers, modulation units, standard sensing fibers, photodetectors, and PCIE6920 acquisition cards, etc., without the need for customized special optical paths and devices; the software algorithm can be implemented on general-purpose CPUs or GPUs, with low development and deployment costs, and can be easily integrated into existing monitoring platforms for linear infrastructure such as pipelines and power cables, showing good prospects for engineering applications. Detailed Implementation A Distributed Fiber Optic Vibration Signal Demodulation and Localization Method Based on Intelligent Optimization Algorithm Example 1: A 30 km pipeline intrusion monitoring system based on PCIE6920 This embodiment establishes a distributed fiber optic vibration monitoring system suitable for intrusion monitoring of a 30km oil pipeline. The hardware consists of a narrow-linewidth continuous-wave laser, an acousto-optic modulator, single-mode sensing fibers laid along the pipeline, a photodetector, a PCIE6920 high-speed data acquisition card, an industrial computer, and an alarm and display terminal. The PCIE6920 data acquisition card operates in single-channel I / Q demodulation mode with a sampling rate of 250MSa / s. The number of sampling points and data upload rate can be flexibly configured according to the actual monitoring distance and spatial resolution requirements. The system's software processing flow is deployed on an industrial computer. First, it acquires real-time I / Q channel data from a PCIE6920 acquisition card, synthesizes it into a complex signal, and then calculates the amplitude sequence and phase sequence distributed along the optical fiber using the complex signal, providing a basis for subsequent signal processing. To effectively suppress random noise while highlighting amplitude changes caused by vibration, this embodiment employs an amplitude demodulation algorithm combining moving average and moving difference. After processing the amplitude sequence, the amplitude change is obtained and compared with a preset vibration detection threshold to filter out a set of suspected vibration locations. For the phase sequence, spatial decoupling and temporal decoupling are performed sequentially to restore continuous phase changes. For intrusion events involving high-frequency vibration and drastic phase changes on pipelines, a downsampling difference decoupling method is used to process the phase sequence, reducing computational load while ensuring the stability of phase decoupling. After phase dewinding, the minimum and mean values ​​of the amplitude sequence within each set distance window are calculated. Coherent fading intervals are identified using a dual-threshold criterion. Interpolation fitting is then used to correct erroneous phases within these intervals, eliminating interference from coherent fading. Finally, within the set of suspected vibration locations, the corrected phase sequence is digitally filtered and denoised to extract the vibration waveforms at each location, thus achieving precise location and waveform recovery of vibration events within a 30km pipeline. Actual experimental results show that the monitoring system built in this embodiment can achieve a spatial positioning resolution of better than 10m within a 30km oil pipeline range, and its frequency response range fully covers 0-10kHz. It can accurately identify various pipeline intrusion behaviors such as digging, impact, and climbing, and fully meet the engineering application requirements for long-distance pipeline intrusion monitoring. Example 2: Construction of Demodulation Parameter Vector and Objective Function This embodiment focuses on achieving unified modeling of demodulation parameters and constructing an optimization objective function, laying the foundation for parameter optimization in the Bat Algorithm. First, it identifies key control parameters throughout the entire process of amplitude demodulation, phase dewinding, and coherent fading suppression. Parameters such as amplitude demodulation window length, spatial moving average window length, phase downsampling factor, differential order, core filter parameters, coherent fading dual-judgment threshold, and vibration detection threshold are integrated to construct a multi-dimensional demodulation parameter vector, enabling unified management of all key control parameters. To objectively evaluate the demodulation performance corresponding to different parameter vectors, calibration data containing multiple typical pipeline intrusion vibration events were selected as the test set. Based on this test set, four core evaluation indicators were calculated: first, the average positioning error, which reflects the accuracy of vibration location positioning; second, the vibration waveform correlation coefficient, which reflects the similarity between the demodulated waveform and the real waveform; third, the signal-to-noise ratio improvement, which reflects the system's ability to suppress background noise; and fourth, the single-frame processing time, which reflects the system's real-time processing performance. The four evaluation indicators mentioned above are normalized to eliminate differences in dimensions and numerical ranges. Then, based on the actual needs of engineering applications, preset weights are assigned to each indicator. The normalized indicators are then linearly combined to construct an objective function with the demodulation parameter vector as the variable. The objective function is optimized by minimizing the parameter vector; a smaller value indicates better overall performance of the corresponding parameter vector in terms of positioning accuracy, waveform reconstruction, noise suppression, and real-time performance. This sets a clear and quantifiable optimization objective for the subsequent global optimization of the bat algorithm. Example 3: Parameter Optimization and Online Update Based on Bat Algorithm This embodiment details the demodulation parameter optimization process based on the bat algorithm and completes the online update and configuration of system parameters, which is divided into two stages: offline optimization and online update. Offline parameter optimization First, complete the hyperparameter initialization of the bat algorithm, setting the number of individual bats, maximum number of iterations, frequency variation range, initial loudness, initial emission probability, etc. Within the parameter range that conforms to physical meaning and engineering practice, randomly generate the initial parameter vector position and velocity corresponding to each individual bat. Entering the iterative optimization phase, in each iteration, a random frequency is first assigned to each individual according to the rules. The velocity and position of each individual are then updated based on the global optimal solution, ensuring that the parameter vector remains within a reasonable range. A complete signal demodulation process is then executed based on the updated parameter vector. The objective function value is calculated based on the calibration data, and the objective function values ​​of all individuals are compared to update the global optimal parameter vector. To improve optimization accuracy, a local perturbation mechanism is introduced. When the random number is greater than the individual's transmission probability, a local perturbation is performed on the individual's position, centered on the current global optimal parameter vector, thus combining global and local search. Then, a new solution is determined. If the random number is less than the individual loudness and the objective function value of the new parameter vector after perturbation is better than the original parameter vector, the new solution is accepted, and the individual loudness is increased and the emission probability is decreased according to the rules. If the determination conditions are not met, the original parameter vector is retained. After each round of iteration, it is determined whether the maximum number of iterations has been reached or the change in the objective function is less than a preset threshold. If both are met, the iteration is terminated, and the globally optimal demodulation parameter vector is output. If not, the iteration continues until convergence is achieved. Online parameter updates The optimal parameter vector obtained through offline optimization is solidified as the initial system configuration. During the actual pipeline monitoring phase, the system performs real-time signal demodulation and positioning according to this parameter vector. In the long-term operation of the system, a parameter adaptive update mechanism is established: when the system detects a significant change in the statistical characteristics of environmental noise, or when the continuous operating time reaches a preset period, it automatically enters a re-optimization mode. The system collects recent monitoring data, including typical vibration events and environmental noise, in real time. This data is used as a new training set to re-execute the objective function calculation and the bat algorithm iterative optimization process, obtaining a new optimal parameter vector adapted to the current operating conditions and environment. The system configuration is then automatically updated, replacing the original parameters. This online update mechanism enables the system to adapt to changes in the environment and pipeline operating conditions in real time, maintaining optimal demodulation and positioning performance and effectively avoiding performance degradation caused by environmental changes. Attached Figure Description Figure 1 This is a comparison chart of key parameter indicators in the pipeline monitoring system before and after optimization using this method.

Claims

1. A distributed optical fiber vibration signal demodulation and localization method based on intelligent optimization algorithm, characterized in that, The method is applied to a phase-sensitive optical time-domain reflectometer distributed fiber optic vibration monitoring system, and includes the following steps: 1) Transmitting optical detection signals and receiving backscattered Rayleigh signals: A narrow-linewidth laser outputs continuous light, which is modulated by a modulation unit into a detection light pulse with a preset pulse width and repetition frequency. The detection light pulse includes a single-frequency pulse, a compressed pulse, and / or a multi-band chirped pulse. The detection light pulse propagates along the distributed optical fiber and generates backscattered Rayleigh light, which is converted into an electrical signal by a photodetector. 2) I / Q demodulation and data acquisition: The electrical signal is beat-frequency with the local oscillator to obtain the intermediate frequency signal; the intermediate frequency signal is sampled by a high-speed sampling board, and the I-channel and Q-channel data distributed along the fiber length direction are obtained by using the built-in digital down-conversion and I / Q demodulation modules of the board. 3) Amplitude demodulation and coarse vibration localization: Calculate the amplitude sequence based on I / Q data; perform direct difference, continuous averaging and difference and / or moving average and moving difference on the amplitude sequences of different scan cycles to obtain the amplitude change distribution, and compare the amplitude change with a preset threshold to obtain the suspected vibration location range; 4) Phase demodulation and vibration waveform recovery: Calculate the phase sequence based on the I / Q data, and sequentially perform range expansion, subtraction of the initial phase, spatial unwinding, and temporal unwinding on the phase sequence. When the data volume is large or the vibration frequency is high, perform downsampling unwinding and / or downsampling differential unwinding on the phase sequence, and then map it back to the original sampling points. Filter and denoise the unwound phase sequence within the vibration position range to obtain the vibration waveform at the target position. 5) Coherent fading suppression: Identify amplitude trough intervals based on local statistics of the amplitude sequence, and correct the phase value within the trough intervals through neighborhood interpolation, frequency domain decomposition and / or multi-band synthesis to eliminate pseudo-phase caused by coherent fading; 6) Intelligent optimization modeling and parameter optimization: The differential window length, downsampling factor, filter parameters and threshold values ​​involved in amplitude demodulation, phase demodulation and coherent fading suppression are composed into a parameter vector. The objective function is constructed with positioning error, vibration waveform reconstruction error, signal-to-noise ratio and processing time. The bat algorithm is used to iteratively optimize the parameter vector until the preset number of iterations or the objective function converges, and the optimal demodulation parameters are obtained. 7) Real-time monitoring: Perform steps 2) to 5) above according to the optimal demodulation parameters to realize real-time demodulation and positioning of distributed optical fiber vibration signals.

2. The method as described in claim 1, characterized in that, Step 3) involves amplitude demodulation and coarse vibration localization, which includes at least one of the following methods: (1) Direct difference: Subtract the amplitude data of two or more adjacent frames point by point according to the distance to obtain the amplitude difference sequence, and determine the vibration interval by comparing the absolute value of the difference with the threshold; (2) Continuous averaging and differencing: Averaging the amplitude data of multiple consecutive frames by frame number to suppress random noise, and then performing inter-frame or intra-frame differencing on the averaging result; (3) Moving average and moving difference: The amplitude sequence is averaged in the distance direction and then the moving average result is differentiald in the time direction to obtain an amplitude change that is more robust to low signal-to-noise ratio environments.

3. The method as described in claim 1, characterized in that, Step 4) includes phase demodulation and vibration waveform recovery, which includes: (1) Range expansion: Perform linear or nonlinear mapping on the original phase sequence to expand the phase value from (-π,π] to a larger numerical range; (2) Subtract the initial phase: Select the phase of the unperturbed reference time or reference spatial position as a reference, and subtract the reference phase from the subsequent data to eliminate static bias; (3) Spatial unwinding: Detect phase jumps along the fiber distance direction and eliminate 2π periodic jumps in the spatial direction by adding or subtracting 2π. (4) Time unwinding: Perform phase unwinding on the time-varying phase sequence at a fixed distance position to eliminate the 2π periodic jump in the time direction; (5) Differential dewinding: The phase sequence is de-differentiated in time or space before dewinding to reduce the impact of low-frequency drift on the dewinding result; (6) Downsampling unwinding and downsampling differential unwinding: When the phase change is drastic or the data volume is large, the phase sequence is downsampled by a predetermined multiple, and unwinding or differential unwinding is performed on the downsampled sequence. Then, the original sampling points are mapped back by interpolation or sample copying.

4. The method as described in claim 1, characterized in that, Step 5) specifically includes coherent fading detection and suppression: (1) Calculate the minimum and mean values ​​of the amplitude sequence within each distance window. When the minimum value is less than the first threshold and the mean value is less than the second threshold, mark the window as a coherent fading interval. (2) Perform polynomial fitting, regularized interpolation, or vector rotation based on the reference phase on the non-fading phase sequences on both sides of the marked coherent fading interval to obtain continuous phase curves; (3) Replace the phase value in the coherent fading interval with the phase value corresponding to the continuous phase curve, or use the rotation vector sum algorithm of the multi-band scattering signal to synthesize the amplitude and phase, thereby reducing the influence of the erroneous phase on the vibration waveform.

5. The method as described in claim 1, characterized in that, The construction of the objective function J(θ) in step 6) includes: (1) Calculate the average positioning error between the vibration positioning result obtained from the current parameter vector and the actual intrusion location, and normalize it into the first index; (2) Calculate the correlation coefficient between the demodulated vibration waveform and the reference sensor or high-precision offline demodulation result, and normalize its negative value or 1-correlation coefficient as the second index; (3) Calculate the increase in signal-to-noise ratio before and after demodulation, and normalize its negative value to the third index; (4) Calculate the processing time required for a single frame of data from acquisition to output vibration waveform, and normalize it into a fourth index; (5) The objective function J(θ) is obtained by linearly combining the above indicators according to the preset weights. The Bat Algorithm takes minimizing J(θ) as the optimization objective.

6. The method as described in claim 1, characterized in that, The iterative process of the Bat Algorithm includes: Step 1: Initialize the number of individual bats, the maximum number of iterations, the frequency range, the initial loudness and the emission probability, and randomly generate the initial parameter vector position and velocity of each individual; Step 2: Update the velocity and position of each individual according to the current frequency, and restrict the parameter vector within the preset physical meaning range; Step 3: Execute the demodulation process based on the updated parameter vector, calculate the objective function value, and update the global optimal parameter vector; Step 4: When the random number is greater than the individual's emission probability, the individual's position is locally perturbed with the current global optimal parameter vector as the center. Step 5: When the random number is less than the individual loudness and the objective function value of the new solution is better than the old solution, accept the new solution and increase the loudness of the individual and decrease the emission probability. Step 6: Determine whether the maximum number of iterations has been reached or the change in the objective function is less than the threshold. If so, output the current globally optimal parameter vector; otherwise, return to step 2 to continue iterating.

7. The method as described in claim 1, characterized in that, The effective monitoring distance of the sensing fiber is no less than 30 km. Under optimized demodulation parameters, it can achieve vibration localization with a spatial resolution better than 10 m, and the frequency response range covers 0 to 10 kHz.

8. The method as described in claim 1, characterized in that, The high-speed sampling board is a PCIE6920 acquisition card or an equivalent board with a sampling rate of not less than 250 MSa / s and I / Q demodulation function. The acquisition card works in single-channel demodulation mode.

9. The method as described in claim 1, characterized in that, In the real-time monitoring and online adjustment steps, during the normal monitoring phase, the system automatically enters the re-optimization mode at a fixed period or when a change in the statistical characteristics of environmental noise is detected. It collects a data set containing typical vibration events and noise as a training set, and re-executes the objective function calculation and bat algorithm iteration to update the demodulation parameters.

10. A distributed optical fiber vibration monitoring system based on an intelligent optimization algorithm, characterized in that, include: Light source and modulation unit, sensing fiber optic, photoelectric detection unit, high-speed sampling board, data processing unit and alarm and display unit; The light source and modulation unit are used to generate a detection light pulse with a preset pulse width and repetition frequency; the sensing fiber is used to be laid along the monitored pipeline and generate a backscattered Rayleigh signal coupled with external vibration; the photoelectric detection unit is used to convert the backscattered Rayleigh signal into an electrical signal; the high-speed sampling board is used to sample the intermediate frequency signal and output I / Q data; the data processing unit is used to execute the method described in any one of claims 1 to 9 to realize the demodulation and localization of the distributed vibration signal; the alarm and display unit is used to output alarm information when an intrusion event is detected, and to mark the vibration location along the pipeline, the event type, and the time information.

11. The system as claimed in claim 10, characterized in that, The data processing unit is an industrial computer or an FPGA and CPU co-processing platform. Acquisition control and data buffering are performed by the FPGA, while amplitude demodulation and phase dewinding operations are performed by the CPU or GPU. The bat algorithm runs on the CPU or GPU to meet a system response time of no more than 0.5 s.

Images