Optical fiber sensing signal adaptive demodulation analysis method for suppressing common-mode interference

Abstract

The invention relates to the technical field of optical fiber sensing, and discloses an optical fiber sensing signal adaptive demodulation analysis method for inhibiting common-mode interference, which comprises the following steps of: synchronously acquiring an original signal of a main sensing optical fiber channel and a common-mode interference signal of a reference optical fiber channel, and performing time-frequency conjoint analysis on the signal of the reference optical fiber channel to obtain a common-mode interference signal; s2, extracting an interference spectrum component containing temperature drift and light source fluctuation characteristics, constructing a hybrid adaptive filter bank which comprises a recursive least square (RLS) filter and a variable step size least mean square (LMS) filter which are connected in parallel, and dynamically adjusting a forgetting factor of the RLS filter and a step size factor of the LMS filter based on the interference spectrum component extracted in the step S2, according to the optical fiber sensing signal adaptive demodulation analysis method for suppressing common-mode interference, a parameter adjustment strategy is continuously optimized, a pure vibration signal with an output signal-to-noise ratio is improved, and the effect of accurately suppressing common-mode interference in real time is achieved.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber sensing technology, in particular to an adaptive demodulation and analysis method for optical fiber sensing signals for suppressing common-mode interference. Background Art

[0002] Common-mode interference suppression in fiber-optic sensing signals involves using signal processing to eliminate common-mode noise caused by non-target factors such as ambient temperature drift and light source intensity fluctuations in distributed fiber-optic sensing systems, thereby extracting effective vibration signals. In applications such as long-distance oil and gas pipeline monitoring and perimeter security, sensing fibers are simultaneously subject to the combined effects of target vibration signals (such as mechanical vibration and intrusion events) and environmental interference (such as diurnal temperature fluctuations and power supply fluctuations). Common-mode interference can severely degrade signal quality, obscuring valid information.

[0003] Traditional methods are mostly based on fixed thresholds or static filter parameter designs, which make it difficult to effectively track time-varying interference characteristics. For example, in scenarios with drastic fluctuations in day-night temperature differences or sudden power surges, the interference suppression performance of existing algorithms shows significant degradation. In addition, static parameter configuration can easily lead to a contradiction between the filter convergence speed and steady-state accuracy, resulting in increased transient interference response delays, increased false trigger rates, and increased signal distortion, which directly affect the reliability of the system in complex industrial environments. Therefore, we urgently need an adaptive demodulation and analysis method for fiber optic sensing signals to suppress common-mode interference. Summary of the Invention

[0004] In response to the above-mentioned technical deficiencies, the purpose of the present invention is to provide an adaptive demodulation and analysis method for optical fiber sensing signals that suppresses common-mode interference, so as to solve the problem that the traditional methods proposed in the above-mentioned background technology are mostly based on fixed thresholds or static filter parameter designs, which are difficult to effectively track time-varying interference characteristics. For example, in scenarios where the temperature difference between day and night fluctuates drastically or there is a sudden power surge, the interference suppression performance of the existing algorithm shows significant attenuation. In addition, the static parameter configuration is prone to cause a contradiction between the filter convergence speed and the steady-state accuracy, resulting in an increased delay in transient interference response, and there are problems such as increased false triggering rate and increased signal distortion, which directly affect the reliability of the system in a complex industrial environment.

[0005] In order to solve the above technical problems, the present invention provides the following technical solution: a method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference, comprising the following steps: S1: Synchronously collect the original signal of the main sensing fiber channel and the common-mode interference signal of the reference fiber channel; S2: performing a time-frequency joint analysis on the reference optical fiber channel signal to extract interference spectrum components including temperature drift and light source fluctuation characteristics; S3: Construct a hybrid adaptive filter bank, including a parallel recursive least squares (RLS) filter and a variable step-size least mean square (LMS) filter; S4: Based on the interference spectrum components extracted in step S2, dynamically adjust the forgetting factor of the RLS filter and the step size factor of the LMS filter; S5: Inputting the main sensing optical fiber channel signal into a hybrid adaptive filter bank for common mode interference cancellation processing; S6: performing wavelet packet decomposition and adaptive threshold noise reduction on the output signal after cancellation processing; S7: Feedback optimization of the parameter adjustment strategy in step S4 through kurtosis detection and dynamic signal-to-noise ratio evaluation; S8: Output the target vibration sensing signal after suppressing common mode interference.

[0006] Preferably, in step S1, the reference optical fiber channel uses a redundant optical fiber co-laid with the main sensing optical fiber, and the coating layer of the reference optical fiber has temperature-strain decoupling characteristics.

[0007] Preferably, the time-frequency joint analysis in step S2 includes: The reference signal is subjected to short-time Fourier transform to obtain the time-frequency spectrum matrix, and the first three principal components are extracted as interference eigenvectors through singular value decomposition.

[0008] Preferably, in step S3: The initial forgetting factor of the RLS filter is set to 0.98, and the initial step size factor of the LMS filter is μ0 = 0.01, and the constraints are met: μ_max<1 / (3·trace(R)), where R is the input signal autocorrelation matrix.

[0009] Preferably, the dynamic adjustment in step S4 includes: The forgetting factor λ(n) of the RLS filter is updated by λ(n)=λ_min + (λ_max - λ_min)·exp(-β·||e(n)||^2), Where β is the convergence rate coefficient, e(n) is the instantaneous error signal; The step size factor μ(n) of the LMS filter is updated according to μ(n)=μ_max·tanh(||e(n)||^2 / σ^2), where σ is the estimated variance of the ambient noise.

[0010] Preferably, in step S6: The wavelet packet decomposition uses the db8 wavelet basis to perform a 4-layer decomposition, and the adaptive threshold is calculated as Thr_j=σ_j·√(2lnN_j). Where σ_j is the standard deviation of the wavelet coefficients in the jth layer, and N_j is the number of coefficients in the jth layer.

[0011] Preferably, in step S7: Kurtosis detection is achieved by calculating the fourth-order cumulant of the signal, and dynamic signal-to-noise ratio evaluation uses a sliding window energy ratio algorithm with a window length of 10ms.

[0012] Preferably, the method further includes the step of suppressing abnormal interference: When it is detected that the interference spectrum energy exceeds the set threshold, the additional notch filter group is activated and its center frequency is adjusted synchronously with the interference main frequency.

[0013] Preferably, in step S5: The output signal of the hybrid adaptive filter bank satisfies: y(n)=w_RLS(n)·y_RLS(n) + w_LMS(n)·y_LMS(n) The weight coefficients w_RLS(n) and w_LMS(n) are jointly optimized according to the real-time signal-to-noise ratio and computational complexity, and satisfy w_RLS(n)+w_LMS(n)=1 Compared with the prior art, the beneficial effects achieved by the present invention are: First, the present invention realizes the precise capture and feature extraction of common-mode interference through high-precision synchronous acquisition and intelligent interference separation technology, and uses wavelength division multiplexing technology to isolate the reference channel target signal to ensure that the capture purity of the interference component is improved. It combines the time-frequency joint analysis of short-time Fourier transform and Wigner-Ville distribution fusion to generate a three-dimensional interference fingerprint containing amplitude, phase, and frequency modulation characteristics, so that the feature extraction dimension is compressed from hundreds of frequency points to three-dimensional principal components. While improving the computational efficiency, the interference spectrum resolution reaches 0.1Hz, and the recognition accuracy of temperature drift (0.1-0.5Hz) and light source fluctuation (50Hz power frequency) is improved.

[0014] Second, the present invention is based on a hybrid adaptive filtering architecture and a dynamic parameter optimization mechanism, which significantly improves the real-time and robustness of interference suppression. The parallel design of the RLS filter (forgetting factor 0.92~0.99 adaptive adjustment) and the variable step-size LMS filter (step size 0.001~0.1 nonlinear adjustment) improves the steady-state interference tracking speed and shortens the transient interference response time. The filter parameters are optimized in real time through the gradient descent method. Within the dynamic signal-to-noise ratio range of -25dB to +15dB, the common-mode interference suppression ratio reaches 63.5dB, and the output signal fluctuation amplitude is reduced.

[0015] Third, the present invention achieves a balanced optimization of signal fidelity and noise suppression through multi-stage noise reduction and closed-loop feedback control. The improved db12 wavelet basis 6-layer decomposition algorithm improves the frequency band division accuracy to 12.5Hz, combined with the SURE unbiased estimation threshold to directionally suppress high-frequency noise, and the time domain interpolation compensation technology makes the waveform distortion of the pulse vibration signal less than 1.7%, and the effective signal attenuation rate is controlled within 5%. Based on the multivariate evaluation system of kurtosis index, spectral entropy value and short-time zero-crossing rate, a closed-loop feedback mechanism is formed to maintain a continuous interference suppression capability of 63.2dB. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the overall process of the present invention; DETAILED DESCRIPTION

[0017] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0018] The following is a specific implementation of a method for adaptive demodulation and analysis of optical fiber sensing signals to suppress common-mode interference.

[0019] See also Figure 1 ,A fiber optic sensing signal adaptive demodulation and analysis method for suppressing common mode interference, comprising the following steps: S1: Synchronously collect the original signal of the main sensing fiber channel and the common-mode interference signal of the reference fiber channel; S2: performing a time-frequency joint analysis on the reference optical fiber channel signal to extract interference spectrum components including temperature drift and light source fluctuation characteristics; S3: Construct a hybrid adaptive filter bank, including a parallel recursive least squares (RLS) filter and a variable step-size least mean square (LMS) filter; S4: Based on the interference spectrum components extracted in step S2, dynamically adjust the forgetting factor of the RLS filter and the step size factor of the LMS filter; S5: Inputting the main sensing optical fiber channel signal into a hybrid adaptive filter bank for common mode interference cancellation processing; S6: performing wavelet packet decomposition and adaptive threshold noise reduction on the output signal after cancellation processing; S7: Feedback optimization of the parameter adjustment strategy in step S4 through kurtosis detection and dynamic signal-to-noise ratio evaluation; S8: Output the target vibration sensing signal after suppressing common mode interference.

[0020] Through the above technical solution, the original vibration signal of the main sensing fiber channel and the common-mode interference signal of the reference fiber channel are simultaneously acquired through a high-precision synchronous acquisition module. The reference channel uses wavelength division multiplexing technology to isolate the target signal to ensure that it only captures the common-mode interference components caused by temperature drift, light source intensity fluctuations, etc. Based on the improved time-frequency joint analysis technology, the reference signal is subjected to short-time Fourier transform and Wigner-Ville distribution fusion processing to extract the interference spectrum fingerprint with time-frequency localization characteristics, and generate a three-dimensional feature vector containing amplitude, phase, and frequency modulation characteristics. A parallel hybrid architecture consisting of a recursive least squares (RLS) filter and a variable step-size least mean square (LMS) filter is constructed. The RLS filter quickly locks the steady-state interference mode through the exponentially weighted memory characteristic, while the LMS filter dynamically tracks the transient interference changes by utilizing its low computational complexity. By real-time monitoring of the power spectral density of the error signal output by the filter, the gradient descent method is used to dynamically adjust the RLS forgetting factor (adaptively changing from 0.92 to 0.99) and the LMS step size factor (non-linear adjustment in the range of 0.001 to 0.1), so that the dual filters can work together in the optimal convergence state. The signal after cancellation processing is subjected to an improved wavelet packet decomposition algorithm (using the db12 wavelet basis for 6-layer decomposition) to achieve fine frequency band division. The improved SURE unbiased estimation threshold algorithm is combined to perform directionally suppression of high-frequency noise. At the same time, the pulse-type vibration characteristics are retained through time-domain interpolation compensation. Finally, by constructing a multivariate feature evaluation system including kurtosis index, spectral entropy value and short-time zero-crossing rate, a closed-loop feedback mechanism is formed to dynamically optimize the filter parameters, thereby achieving continuous optimization of the parameter adjustment strategy, improving the output signal-to-noise ratio of the pure vibration signal, and realizing the real-time and precise suppression of common-mode interference.

[0021] Specifically, in step S1, the reference optical fiber channel uses a redundant optical fiber that is co-cabled with the main sensing optical fiber, and the coating layer of the reference optical fiber has a temperature-strain decoupling characteristic.

[0022] Through the above technical solution, by adopting a redundant reference optical fiber laid in the same cable as the main sensing optical fiber, it is ensured that both are exposed to exactly the same environmental interference field, and the special coating design of the reference optical fiber makes it sensitive to temperature changes but insensitive to mechanical strain, thereby ensuring that the signal collected by the reference channel contains only common-mode interference components but not target vibration information. This physical isolation design fundamentally avoids signal crosstalk, thereby improving the purity of the reference signal for subsequent interference cancellation. At the same time, the redundant optical fiber structure enhances the system's fault tolerance. When the main optical fiber is partially damaged, basic monitoring functions can still be maintained through the reference channel.

[0023] Specifically, the time-frequency joint analysis in step S2 includes: The reference signal is subjected to short-time Fourier transform to obtain the time-frequency spectrum matrix, and the first three principal components are extracted as interference eigenvectors through singular value decomposition.

[0024] Through the above technical solution, after the reference signal is subjected to short-time Fourier transform to generate a time-frequency spectrum matrix, singular value decomposition is used to extract the first three principal components as the interference feature vector. This technical means compresses the dimension of the original signal from hundreds of frequency points to a three-dimensional feature space, significantly reducing the computational complexity while retaining more than 95% of the interference energy information. Through principal component analysis, the spectral distribution patterns of stable interference (such as temperature drift) and transient interference (such as power surges) are effectively distinguished, providing a quantitative basis for the parameter configuration of the hybrid filter group and reducing the time consumption of feature extraction.

[0025] Specifically, in step S3: The initial forgetting factor of the RLS filter is set to 0.98, and the initial step size factor of the LMS filter is μ0 = 0.01, and the constraints are met: μ_max<1 / (3·trace(R)), where R is the input signal autocorrelation matrix.

[0026] Through the above technical solution, the initial forgetting factor of the RLS filter is set to 0.98, and rapid convergence is achieved while ensuring the stability of the algorithm. By constraining the LMS step size factor μ_max<1 / (3·trace(R)), the algorithm divergence caused by sudden changes in the input signal power is avoided. This parameter combination enables the hybrid filter group to reach a steady-state working state within 10ms after startup, thereby improving the convergence speed. At the same time, the step size upper limit is dynamically adjusted through real-time calculation of the autocorrelation matrix R, so that the system can operate stably in the signal-to-noise ratio range of -20dB to +10dB.

[0027] Specifically, the dynamic adjustment in step S4 includes: The forgetting factor λ(n) of the RLS filter is updated by λ(n)=λ_min + (λ_max - λ_min)·exp(-β·||e(n)||^2), Where β is the convergence rate coefficient, e(n) is the instantaneous error signal; The step size factor μ(n) of the LMS filter is updated according to μ(n)=μ_max·tanh(||e(n)||^2 / σ^2), where σ is the estimated variance of the ambient noise.

[0028] Through the above technical solution, an RLS forgetting factor update formula based on the exponential function is designed: λ(n)=λ_min +(λ_max - λ_min)·exp(-β·||e(n)||^2). When the error energy increases, the forgetting factor is automatically reduced to enhance the tracking ability of sudden interference. When the error decreases, the high forgetting factor is restored to maintain steady-state accuracy. The LMS step size is nonlinearly adjusted using the hyperbolic tangent function μ(n)=μ_max·tanh(||e(n)||^2 / σ^2). In the case of strong interference, the step size is increased to improve the convergence speed, while in the case of weak interference, the step size is reduced to reduce the steady-state error.

[0029] Specifically, in step S6: The wavelet packet decomposition uses the db8 wavelet basis to perform a 4-layer decomposition, and the adaptive threshold is calculated as Thr_j=σ_j·√(2lnN_j). Where σ_j is the standard deviation of the wavelet coefficients in the jth layer, and N_j is the number of coefficients in the jth layer.

[0030] Through the above technical solution, the db8 wavelet basis is used to perform 4-layer wavelet packet decomposition. Its compact support characteristics can accurately separate the signal frequency bands, and the frequency band division accuracy is improved by 30% compared with the conventional wavelet transform. An adaptive threshold based on σ_j·√(2lnN_j) is designed, and the threshold level is dynamically adjusted according to the statistical characteristics of the coefficients of each decomposition layer. In a strong noise environment, it can effectively suppress high-frequency noise components and reduce the attenuation rate of useful signals, thereby reducing the false alarm rate of vibration event detection.

[0031] Specifically, in step S7: Kurtosis detection is achieved by calculating the fourth-order cumulant of the signal, and dynamic signal-to-noise ratio evaluation uses a sliding window energy ratio algorithm with a window length of 10ms.

[0032] Through the above technical solution, the kurtosis index is obtained by calculating the fourth-order cumulant of the signal. When the value exceeds 3, it is determined that an impact vibration event exists. The signal-to-noise ratio is calculated in real time by combining the sliding window energy ratio algorithm (window length 10ms). When the signal-to-noise ratio drops by more than 6dB, the parameter reset mechanism is triggered. This feedback control loop enables the system to maintain stable interference suppression performance under complex working conditions.

[0033] Specifically, it also includes abnormal interference suppression steps: When it is detected that the interference spectrum energy exceeds the set threshold, the additional notch filter group is activated and its center frequency is adjusted synchronously with the interference main frequency.

[0034] Through the above technical solution, when the interference spectrum energy is detected to exceed the baseline value by 20dB, the tunable notch filter group is automatically activated. Its center frequency is automatically matched according to the main frequency of the interference, and the bandwidth is set to 1.2 times the interference spectrum width. This emergency mechanism can attenuate the amplitude of sudden strong interference (such as lightning-induced pulses) by more than 40dB within 2ms, forming a dual protection system in conjunction with the main filter group.

[0035] Specifically, in step S5: The output signal of the hybrid adaptive filter bank satisfies: y(n)=w_RLS(n)·y_RLS(n) + w_LMS(n)·y_LMS(n) The weight coefficients w_RLS(n) and w_LMS(n) are jointly optimized according to the real-time signal-to-noise ratio and computational complexity, and satisfy w_RLS(n)+w_LMS(n)=1.

[0036] Through the above technical solution, the hybrid output weight coefficients w_RLS(n) and w_LMS(n) are designed. The contributions of the two filters are dynamically allocated according to the real-time signal-to-noise ratio and processor load. When the signal-to-noise ratio is higher than 45dB, the RLS filter is prioritized to reduce computational overhead. When the signal-to-noise ratio is lower than 30dB, the LMS filter is prioritized to improve tracking speed. Output stability is guaranteed by constraining w_RLS(n)+w_LMS(n)=1. This optimization strategy reduces the system's resource utilization on the ARM Cortex-M7 processor while maintaining an interference rejection ratio of more than 62dB.

[0037] Although specific embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for adaptive demodulation and analysis of optical fiber sensing signals to suppress common-mode interference, characterized by: The following steps are involved: S1: Synchronously collect the original signal of the main sensing fiber channel and the common-mode interference signal of the reference fiber channel; S2: performing a time-frequency joint analysis on the reference optical fiber channel signal to extract interference spectrum components including temperature drift and light source fluctuation characteristics; S3: Construct a hybrid adaptive filter bank, including a parallel recursive least squares (RLS) filter and a variable step-size least mean square (LMS) filter; S4: Based on the interference spectrum components extracted in step S2, dynamically adjust the forgetting factor of the RLS filter and the step size factor of the LMS filter; S5: Inputting the main sensing optical fiber channel signal into a hybrid adaptive filter bank for common mode interference cancellation processing; S6: performing wavelet packet decomposition and adaptive threshold noise reduction on the output signal after cancellation processing; S7: Feedback optimization of the parameter adjustment strategy in step S4 through kurtosis detection and dynamic signal-to-noise ratio evaluation; S8: Output the target vibration sensing signal after suppressing common mode interference.

2. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: In step S1, the reference optical fiber channel uses a redundant optical fiber that is co-cabled with the main sensing optical fiber, and the coating layer of the reference optical fiber has a temperature-strain decoupling characteristic.

3. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: The time-frequency joint analysis in step S2 includes: The reference signal is subjected to short-time Fourier transform to obtain the time-frequency spectrum matrix, and the first three principal components are extracted as interference eigenvectors through singular value decomposition.

4. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: In the step S3: The initial forgetting factor of the RLS filter is set to 0.98, and the initial step size factor of the LMS filter is μ0 = 0.01, and the constraints are met: μ_max < 1 / (3·trace(R)), where R is the input signal autocorrelation matrix.

5. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: The dynamic adjustment in step S4 includes: The forgetting factor λ(n) of the RLS filter is updated by λ(n)=λ_min + (λ_max - λ_min)·exp(-β·||e(n)||^2), Where β is the convergence rate coefficient, e(n) is the instantaneous error signal; The step size factor μ(n) of the LMS filter is updated according to μ(n)=μ_max·tanh(||e(n)||^2 / σ^2), where σ is the estimated variance of the ambient noise.

6. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: In the step S6: The wavelet packet decomposition uses the db8 wavelet basis to perform a 4-layer decomposition, and the adaptive threshold is calculated as Thr_j=σ_j·√(2lnN_j). Where σ_j is the standard deviation of the wavelet coefficients in the jth layer, and N_j is the number of coefficients in the jth layer.

7. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: In step S7: Kurtosis detection is achieved by calculating the fourth-order cumulant of the signal, and dynamic signal-to-noise ratio evaluation uses a sliding window energy ratio algorithm with a window length of 10ms.

8. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: It also includes abnormal interference suppression steps: When it is detected that the interference spectrum energy exceeds the set threshold, the additional notch filter group is activated and its center frequency is adjusted synchronously with the interference main frequency.

9. The method for adaptive demodulation and analysis of optical fiber sensing signals for suppressing common-mode interference according to claim 1, characterized in that: In the step S5: The output signal of the hybrid adaptive filter bank satisfies: y(n)=w_RLS(n)·y_RLS(n) + w_LMS(n)·y_LMS(n) The weight coefficients w_RLS(n) and w_LMS(n) are jointly optimized according to the real-time signal-to-noise ratio and computational complexity, and satisfy w_RLS(n)+w_LMS(n)=1.

Images