Improved moving average method applied to distributed optical fiber vibration sensing system

Abstract

The invention relates to the technical field of optical fiber sensing, and provides an improved moving average method applied to a distributed optical fiber vibration sensing system, and the method comprises the steps: continuously collecting a plurality of backward Rayleigh scattering curves; setting a plurality of different moving average window widths, and performing moving average processing on the plurality of backward Rayleigh scattering curves based on each moving average window width to obtain a plurality of scattering curves after moving average processing corresponding to each moving average window width; for each moving average window width, calculating the variance of the scattering curve after moving average processing at the same time point, and obtaining a variance sequence corresponding to the moving average window width; according to the variance sequence corresponding to each moving average window width, determining the corresponding moving average window width when the variance change tends to be stable, and taking the moving average window width as a target window width; and moving average processing is carried out on a plurality of acquired backward Rayleigh scattering curves by using a target window width, and a disturbance event on the optical fiber is positioned.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensing technology, and in particular to an improved moving average method for distributed fiber optic vibration sensing systems. Background Technology

[0002] Phase-sensitive optical time-domain reflectometry (OTDR), a highly sensitive distributed fiber optic vibration sensing technology, locates weak disturbances by detecting changes in the interference signal of backscattered Rayleigh light. It is widely used in structural health monitoring, pipeline safety early warning, and other fields. In the signal processing stage of phase-sensitive OTDR, the moving average algorithm is one of the most commonly used disturbance localization methods due to its simplicity and ease of implementation. This algorithm averages multiple backscattered Rayleigh curves by setting a window width to suppress noise and extract effective signals. However, the selection of the moving average window width has a decisive impact on the system's positioning accuracy and signal-to-noise ratio: a window that is too small will result in poor noise reduction, while a window that is too large may lead to signal distortion or wasted computational resources. For a long time, the determination of this parameter has lacked objective theoretical basis, mainly relying on the operator's experience or time-consuming trial and error. This not only reduces signal processing efficiency but also makes it difficult to guarantee the optimal window parameter in different application scenarios, thus affecting the overall performance and response speed of the sensing system. Summary of the Invention

[0003] To address the aforementioned shortcomings, the present invention aims to propose a method that monitors the variance decay trend of the backscattering Rayleigh curve under different moving average window widths and uses its stable point as the optimal window criterion to achieve adaptive and rapid selection of the window width parameter, thereby improving signal processing efficiency and ensuring the accuracy and signal-to-noise ratio of disturbance localization.

[0004] To achieve this objective, the present invention adopts the following technical solution: An improved moving average method for distributed fiber optic vibration sensing systems includes the following steps: S1: Continuously collect several backscattering Rayleigh curves, each of which consists of several data points; S2: Set several different moving average window widths, and perform moving average processing on several back Rayleigh scattering curves based on each moving average window width to obtain several moving average processed scattering curves corresponding to each moving average window width; S3: For each moving average window width, calculate the variance of the scattering curve after moving average processing at the same time point to obtain the variance sequence corresponding to the moving average window width. S4: Based on the variance sequence corresponding to each moving average window width, determine the moving average window width corresponding to when the variance change tends to stabilize and use it as the target window width; S5: Use the target window width to perform moving average processing on the collected backscattering Rayleigh curves, and locate the disturbance event on the optical fiber based on the processing result.

[0005] Preferably, in step S1, the data obtained from continuously acquiring multiple backscattering Rayleigh scattering curves satisfy the following relationship: ; in, This represents a data matrix arranged as column vectors. This represents the total number of backscattering Rayleigh scattering curves collected consecutively. Indicates the first Article The backscattering Rayleigh scattering curve signal composed of 10 data points This indicates the number of data points set in a single backscattering Rayleigh scattering curve.

[0006] Preferably, in step S2, the moving average processing based on a given moving average window width satisfies the following relationship: ; in, Indicates the moving average window width. Indicates based on the moving average window width The first one obtained after performing moving average processing Scattering curves, Indicates the first number of the original collection The backscattering Rayleigh scattering curve signal.

[0007] Preferably, in step S3, the first step is to calculate the moving average obtained after performing moving average processing based on the moving average window width. The variance corresponding to the scattering curve Satisfying the relation: ; in, Indicates the moving average window width. Indicates based on the moving average window width The first one obtained after performing moving average processing Scattering curves.

[0008] Preferably, in step S4, the variance change is determined based on the variance sequence corresponding to each moving average window width, and the variance change is determined through the variance change sequence. To measure the variance change sequence Satisfying the relation: ; in, This indicates that when the moving average window width is The corresponding measurement value at that time Indicates the first The variance corresponding to the scattering curve The maximum value.

[0009] Preferably, determining the moving average window width corresponding to when the variance change tends to stabilize and using it as the target window width includes: calculating the change amplitude between adjacent sequence values ​​in the variance change sequence, comparing the change amplitude with a preset stability threshold, and determining that the variance change tends to stabilize when the change amplitude is less than the stability threshold for the first time, and determining the moving average window width corresponding to this time as the target window width.

[0010] Preferably, in step S5, the collected backscattering Rayleigh curves are processed by moving average using the target window width to satisfy the following relationship: ; in, Indicates the use of the target window width For the original collection The first backscattering Rayleigh scattering curve obtained by moving average processing One data value, This represents the target window width. Indicates the first Backscattering Rayleigh scattering curve signal, Indicates by all The complete data sequence after processing with a moving average, arranged in sequence. and All are positive integer indices.

[0011] One of the above technical solutions has the following advantages or beneficial effects: This invention first continuously collects multiple backscattering Rayleigh curves to construct an original dataset. Then, it sets multiple different moving average window widths to process the dataset in parallel, effectively overcoming the limitations of traditional methods that rely on repeated trial and error with a single window. It calculates the variance of the processed curves at the same time point under each window width, constructing a variance sequence reflecting the discrete characteristics of the data, transforming the originally subjective parameter selection problem into an objectively quantifiable mathematical analysis process. Furthermore, it uses the stability of variance changes as the criterion for determining the optimal window. When the variance decays to a stable state as the window width increases, it indicates that further increasing the window has no substantial significance for signal improvement. At this point, the corresponding window width achieves the best balance between noise reduction and signal fidelity. Finally, it uses this adaptively determined target window width to perform moving average processing on the entire dataset, achieving precise location of disturbance events. In summary, by establishing a quantitative correlation model between variance variation and window width, a clear theoretical basis is provided for parameter selection, enabling rapid adaptive determination of window width, which can significantly improve signal processing efficiency. At the same time, the optimal window width ensures the coordination and unity of noise suppression and useful signal preservation, thereby ensuring that the distributed fiber optic vibration sensing system can maintain high-accuracy disturbance localization capability and stable signal-to-noise ratio performance in complex environments. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0013] Figure 1 This is a flowchart of an improved moving average method for distributed fiber optic vibration sensing systems provided in an embodiment of the present invention. Figure 2 This is a structural diagram of a distributed fiber optic sensing system based on a phase-sensitive optical time-domain reflectometer provided in an embodiment of the present invention; Figure 3 This is a graph showing the relationship between the moving average window width and variance provided in an embodiment of the present invention. Detailed Implementation

[0014] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0015] In this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0016] An improved moving average method applied to distributed fiber optic vibration sensing systems, such as... Figure 2 As shown, the distributed fiber optic vibration sensing system is a phase-sensitive optical time-domain reflectometer system, which includes a narrow-linewidth laser, an electro-optic modulator, an erbium-doped fiber amplifier, a sensing fiber, a circulator, and a photodetector, such as... Figure 1 As shown, a preferred embodiment of the present invention includes the following steps: S1: Continuously collect multiple backscattering Rayleigh curves, each of which consists of several data points; It should be noted that the backscattering Rayleigh curve refers to the one-dimensional data sequence formed after photoelectric conversion of backscattered light caused by the microscopic refractive index inhomogeneity of the fiber when pulsed light propagates in the sensing fiber. The horizontal axis of this sequence usually corresponds to the physical position or time delay of the fiber, and the vertical axis is the intensity value of the scattered light. A data point refers to the discrete sampled value that constitutes a single backscattering Rayleigh curve. Each data point corresponds to a specific spatial location on the fiber, and its value reflects the intensity of the backscattered Rayleigh light at that location. Continuous acquisition refers to periodically injecting probe light pulses into the sensing fiber at a fixed pulse repetition frequency, and synchronously recording the complete backscattered Rayleigh signal within each pulse period, forming a set of multiple curves arranged in chronological order. This acquisition process requires a narrow-linewidth laser to ensure coherence. Continuous light is modulated into pulsed light using an acousto-optic modulator or electro-optic modulator. After being injected into the sensing fiber, the scattered signal is received by a high-speed photodetector, discretized and sampled by an analog-to-digital converter, and stored as a digital signal sequence. The acquisition of multiple backscattering Rayleigh curves provides a data basis for subsequent statistical analysis. By comparing the changes in scattered light intensity at the same spatial location under different pulse periods, signal fluctuations caused by external disturbances can be identified.

[0017] S2: Set several different moving average window widths, and perform moving average processing on several back Rayleigh scattering curves based on each moving average window width to obtain several moving average processed scattering curves corresponding to each moving average window width; It should be noted that the moving average window width refers to the number of data frames involved in the averaging calculation in the moving average algorithm. This parameter determines the degree of noise suppression and the temporal resolution of the signal. Window width It is a positive integer representing a consecutive pair of integers. The backscattering Rayleigh curves are arithmetically averaged at their corresponding spatial locations. Moving average processing refers to a mathematical operation that smooths the data sequence. Its core idea is to sum the multiple curves within a window point by point and then divide by the window width to obtain a new average curve. Several backscattering Rayleigh curves refer to a subset of curves selected from the multiple original curves acquired in step S1, according to the window width.

[0018] Understandably, the window width directly determines the average number of iterations. Smaller windows retain more original signal details but have insufficient noise reduction, while larger windows offer significant noise reduction but may erase weak perturbation features. By processing multiple window widths in parallel, we can systematically observe the changes in noise levels as the signal deteriorates. This method increases the variance of the signal and tracks its decay process, thus identifying the critical point where the variance changes from rapid decay to a stable state. This parallel strategy overcomes the limitations of traditional methods that rely on repeated trials within a single window, transforming subjective experience-driven parameter selection into objective data-driven quantitative analysis, which can significantly improve the efficiency and scientific rigor of parameter optimization.

[0019] S3: For each moving average window width, calculate the variance of the scattering curve after moving average processing at the same time point to obtain the variance sequence corresponding to the moving average window width. It should be noted that variance is a statistic that measures the dispersion of a set of data, reflecting the average of the squared deviations of each data value from the arithmetic mean. In step S3, variance is used to quantify the degree of light intensity fluctuation at the same spatial location (i.e., the fiber location corresponding to the same time dimension) of different scattering curves. The same time point refers to the index of the data point corresponding to the same physical location of the fiber in the backscattering Rayleigh curve. The variance sequence refers to the variance sequence for each candidate window width. , and its corresponding The curve after moving average processing is in The variances are calculated at each spatial location, resulting in a variance array of length L. This sequence characterizes the residual fluctuation level that the signal still retains statistically after being smoothed by a specific window.

[0020] Understandably, when the window width is small, the number of curves involved in the averaging is small, and noise is not sufficiently suppressed. Therefore, the values ​​of multiple curves at the same location differ significantly, resulting in a high variance. As the window width increases, the averaging effect strengthens, noise is gradually weakened, data points tend to be more consistent, and variance decreases accordingly. When the window width exceeds a certain threshold, further increasing the number of averaging iterations has a diminishing marginal effect on noise suppression, the rate of variance decay slows significantly, and eventually stabilizes. This can be achieved by calculating different... The variance sequence can capture this pattern of change, providing an objective basis for determining the optimal window.

[0021] S4: Based on the variance sequence corresponding to each moving average window width, determine the moving average window width corresponding to when the variance change tends to stabilize and use it as the target window width; It should be noted that the variance change tending to stabilize means that as the moving average window width increases, the decay rate of the variance sequence gradually changes from significant to weak, and the variance value no longer decreases significantly with the increase of the window, entering a plateau region. The target window width refers to the minimum window width value that satisfies the variance stability criterion; this value is determined as the final moving average parameter used for signal processing. The variance change sequence is a sequence formed by arranging the maximum values ​​of the variance sequences corresponding to different window widths in order of window size, and its elements are... The maximum variance value under the given value. Step S4 identifies the inflection point where the variance decay curve changes from a steep drop to a gentler drop by monitoring the derivative characteristics of the variance decay curve.

[0022] S5: Use the target window width to perform moving average processing on the collected backscattering Rayleigh curves, and locate the disturbance event on the optical fiber based on the processing result.

[0023] It should be noted that disturbance events refer to external physical interference acting on the sensing fiber, such as mechanical vibration, stress changes, temperature fluctuations, or intrusion. These events cause local refractive index changes in the fiber, thereby modulating the phase and intensity of the backscattered Rayleigh light. Location refers to determining the specific spatial position of the disturbance event on the sensing fiber, typically by indexing the curve data points corresponding to the anomaly points in the detected signal, and then converting this into physical distance using the speed of light and sampling timing.

[0024] Understandably, the purpose of step S5 is to apply the optimal parameters determined in the preceding steps to actual signal processing, achieving highly reliable disturbance detection and localization. Using a moving average across all N original curves with the target window width can filter out random noise to the greatest extent possible, while avoiding signal distortion caused by excessive smoothing. The signal-to-noise ratio of the processed data is significantly improved, and signal changes caused by real disturbances are more prominent.

[0025] Preferably, in step S1, the data obtained from continuously acquiring multiple backscattering Rayleigh scattering curves satisfy the following relationship: ; in, This represents a data matrix arranged as column vectors. This represents the total number of backscattering Rayleigh scattering curves collected consecutively. Indicates the first Article The backscattering Rayleigh scattering curve signal composed of 10 data points This indicates the number of data points set in a single backscattering Rayleigh scattering curve.

[0026] It should be noted that the data matrix It is a two-dimensional array structure used to systematically store and manage multiple continuously acquired backscattering Rayleigh scattering curves. Each column of the matrix corresponds to a complete scattering curve, and the column vector arrangement ensures the continuity of spatial location indexes in the vertical dimension, which facilitates subsequent vectorized batch mathematical operations. The column number of the matrix represents the total number of curves acquired at a constant pulse repetition frequency within a fixed time period. Its value directly determines the confidence level and temporal resolution of the statistical analysis. A larger value results in more accurate statistical characteristic estimation, but also increases the computational burden. Backscattering Rayleigh curve signal. It is a collection A one-dimensional array of elements, where each element is a digital value converted from analog to digital, representing the intensity of scattered light at a specific location in the optical fiber. The value of is determined by the length of the sensing fiber, the width of the optical pulse, and the sampling rate of the analog-to-digital converter. Its fixed value ensures the alignment accuracy of different curves in the spatial dimension. The data matrix is ​​composed of column vectors. In terms of physical storage, it can be regarded as The two-dimensional array structure allows data extraction across multiple curves at the same fiber location to be completed through simple column slicing operations, without the need for complex indexes, which greatly improves the efficiency of subsequent parallel processing.

[0027] It is understandable that this is achieved by defining a data matrix. The mathematical expression of this method abstracts continuously acquired discrete scattering data into structured matrix objects, laying the theoretical framework for subsequent vectorized computation and parallel processing. In a phase-sensitive optical time-domain reflectometer system, the raw scattering curves are essentially unstructured time-series streams. Without effective organization, each processing operation requires repeated data reading and alignment, consuming significant computational resources. By arranging N curves as column vectors, the system can transform cross-curve statistical calculations into matrix row or column operations. Utilizing the SIMD instruction set of modern processors or the parallel computing capabilities of GPUs, it can complete batch processing of large-scale data in one go. This data structuring approach not only simplifies algorithm implementation but, more importantly, transforms the traditional moving average algorithm's independent processing of each curve into a holistic processing mode of the data matrix. This enables multi-window parallel computation, fundamentally avoiding the inefficiency of sequential parameter trial in traditional methods and improving the computational efficiency of parameter optimization.

[0028] Preferably, in step S2, the moving average processing based on a given moving average window width satisfies the following relationship: ; in, Indicates the moving average window width. Indicates based on the moving average window width The first one obtained after performing moving average processing Scattering curves, Indicates the first number of the original collection The backscattering Rayleigh scattering curve signal.

[0029] It should be noted that the moving average window width is a positive integer parameter, physically defined as the number of backscattered Rayleigh scattering curves participating in a single averaging operation. This parameter directly determines the smoothing strength and noise suppression capability of the algorithm. Based on the given... Applying a moving average means that the selected... The mathematical process of performing an arithmetic mean operation on the original curve point by point in the spatial dimension, which suppresses random noise by reducing statistical variance. As the first of the original collections The backscattering Rayleigh scattering curve signal is a signal containing A one-dimensional array of the nth data points, representing the nth data point. Complete spatial distribution information acquired from the sensing fiber within one pulse cycle. As the j-th scattering curve after processing, it also has a length of A one-dimensional array, where each element is a... The arithmetic mean of the original curves at the corresponding spatial location represents the position of that spatial location. The statistical average light intensity over a period of time.

[0030] Understandably, moving average processing transforms the abstract concept of average into executable arithmetic operations, resolving the shortcomings of traditional moving average algorithms, such as ambiguous calculation processes and a lack of unified standards in engineering implementation. In a phase-sensitive optical time-domain reflectometry system, the noise component of the backscattering Rayleigh curve follows a statistically independent distribution over multiple pulse periods, while the perturbation signal exhibits time correlation. Through analysis of... The random noise is attenuated to its original value by taking the arithmetic mean of the curves. Meanwhile, relevant signal components are preserved, thereby improving the signal-to-noise ratio.

[0031] Preferably, in step S3, the first step is to calculate the moving average obtained after performing moving average processing based on the moving average window width. The variance corresponding to the scattering curve Satisfying the relation: ; in, Indicates the moving average window width. Indicates based on the moving average window width The first one obtained after performing moving average processing Scattering curves.

[0032] It should be noted that variance It is a statistic used for quantification. The dispersion of the scattering curve after moving average processing at the same spatial location is a measure of the signal's stability and noise suppression effectiveness at that location. A smaller value indicates better consistency among the curves at that point and weaker random noise. In one embodiment, the relationship between the moving average window width and variance is as follows: Figure 3 As shown. Summation term. This represents the squared deviation of each curve from the average curve. The squaring operation amplifies the impact of large deviations, giving outliers a higher weight in variance calculation and thus enabling more sensitive detection of signal anomalies. As a normalization coefficient, it ensures that the variance value does not change with the window width. Linear growth makes different The variances under these conditions are comparable because the coefficient essentially transforms the sum of squared deviations into the mean squared deviations, which conforms to the mathematical definition of variance. In the relational formula... In this context, it specifically refers to the original acquisition of the first... The backscattering Rayleigh curve is used instead of the averaged curve, which requires that the original curve data be retained when calculating the variance for comparison with the average curve in order to evaluate the smoothing effect.

[0033] Understandably, in a phase-sensitive optical time-domain reflectometer system, the core objective of moving average is to suppress random noise through multi-frame averaging, and variance is precisely the ideal metric for measuring this effect: when the window width... When the size is small, the average effect is limited, and the differences between the curves are large. The value is relatively high; with As the noise level increases, it is gradually suppressed, and the curves tend to converge. Monotonically decreasing; when After exceeding a certain threshold, further increasing the window size has a diminishing marginal effect on noise suppression. The rate of decay has slowed significantly. (This was achieved through monitoring...) Follow The changing decay characteristics can identify the inflection point where the variance curve changes from a steep drop to a gentler drop; this inflection point corresponds to... The value is the optimal window width. This is achieved by using a point-by-point calculation method, i.e., for... The variance is calculated for each spatial location to generate a complete variance distribution curve. This allows the system to evaluate the global noise suppression effect and locate variance anomalies. The latter often corresponds to the location of the actual disturbance event, because the disturbance will cause the curve at that location to show a non-noise-related change, which will manifest as a slower variance decay rate or the appearance of a local minimum.

[0034] Preferably, in step S4, the variance change is determined based on the variance sequence corresponding to each moving average window width, and the variance change is determined through the variance change sequence. To measure the variance change sequence Satisfying the relation: ; in, This indicates that when the moving average window width is The corresponding measurement value at that time Indicates the first The variance corresponding to the scattering curve The maximum value.

[0035] It should be noted that the variance variation sequence It is a one-dimensional array arranged in ascending order of the moving average window width, where each element... All correspond to the window width The maximum variance value, along with the variance variation sequence, fully characterizes the limiting effect of the system's noise suppression under different smoothing intensities. (Measurement value) As a single scalar index in the sequence, it simplifies the comparative analysis of performance across different window widths. Its physical meaning represents the performance across the entire sensing fiber range. The noise level at the spatial location with the most severe residual fluctuations after sub-averaging reflects the degree of signal-to-noise ratio improvement of the system under the worst conditions. Function operation refers to the operation on the variance sequence corresponding to the j-th scattering curve. By performing a comprehensive comparison and extracting the maximum value as the representative variance value for that window, this maximum value selection strategy ensures that the parameter optimization process focuses on system performance bottlenecks and avoids masking the risk of positioning failure in local high-noise areas due to averaging evaluation. In this context, the scattering curve specifically refers to the scattering curve after moving average processing. The average curves and their corresponding variance sequences Depend on The variance values ​​at each spatial location are used to filter out the worst-case variance using the max operation, providing a conservative but reliable quantitative benchmark for subsequent stability assessment.

[0036] Understandably, by introducing variance variation sequences... This method compresses multidimensional variance information under different window widths into a one-dimensional monotonic sequence, greatly simplifying the logic for determining the optimal window. In distributed fiber optic sensing systems, noise levels at different spatial locations inherently differ. Some locations may have naturally poor signal-to-noise ratios due to interference fading or polarization mismatch. If average variance or median variance is used to evaluate the overall noise reduction effect, the improvement in these high-noise areas may be overlooked, leading to overly optimistic parameter selection. However, using... Build The sequence forces the algorithm to always focus on the most sensitive spatial points, ensuring that the selected window width provides sufficient noise suppression across the entire fiber optic cable, avoiding false alarms or missed alarms caused by localized high noise. As the window width... Increase The sequence exhibits a monotonically decreasing characteristic, and its decay pattern is directly related to the noise suppression efficiency. When the rate of decrease slows down, it indicates that continuing to increase the window has no substantial effect on improving the signal-to-noise ratio at the worst point. At this time, the corresponding... In order to achieve the optimal value that takes into account the performance of the entire segment, this parameter selection strategy based on extreme value statistics transforms the traditional experience-based global evaluation into a quantifiable and repeatable mathematical judgment, which significantly improves the robustness and practicality of parameter optimization.

[0037] Preferably, determining the moving average window width corresponding to when the variance change tends to stabilize and using it as the target window width includes: calculating the change amplitude between adjacent sequence values ​​in the variance change sequence, comparing the change amplitude with a preset stability threshold, and determining that the variance change tends to stabilize when the change amplitude is less than the stability threshold for the first time, and determining the moving average window width corresponding to this time as the target window width.

[0038] It should be noted that the magnitude of change refers to the variance change sequence. two adjacent elements and The absolute difference or relative rate of change between the two windows quantifies the marginal gain of noise reduction effect brought about by increasing the window width. A larger change indicates that increasing the window is more effective in improving the signal-to-noise ratio, while a smaller change indicates diminishing returns from further window increases. This indicator provides an observable basis for selecting the optimal window. The stability threshold is a pre-set small positive value that defines clear boundary conditions for variance stability. When the change is below this threshold, the system considers that further increasing the window will no longer bring substantial performance improvement. This threshold is usually set comprehensively based on the sensor system's background noise level, dynamic response requirements, and computational resource limitations, and can be adaptively adjusted during system calibration. The first less-than comparison operation means that, starting from the minimum window width, the change in the range between adjacent windows is calculated in ascending order and immediately compared with the threshold. Once the less-than condition is met, the traversal terminates. This strategy ensures that the minimum window width that meets the stability requirements is selected, avoiding signal distortion caused by excessive smoothing. The target window width is finally determined through the above judgment process. The value is used as a parameter for subsequent global moving average processing, and its magnitude directly affects the balance between the signal-to-noise ratio and spatial resolution of the final signal.

[0039] Understandably, this invention aims to transform the optimal window selection problem into a quantifiable threshold determination problem, completely eliminating the inefficient mode of traditional methods that rely on subjective experience or exhaustive trial and error. In distributed fiber optic sensing systems, the variance variation sequence... With window width Increasing the window size exhibits typical nonlinear attenuation characteristics. The initial steep attenuation corresponds to the noise-dominated stage, where further increasing the window size can significantly improve the signal-to-noise ratio. The later, gradual attenuation corresponds to a stage where noise is sufficiently suppressed, and residual fluctuations are mainly caused by inherent system uncertainties or weak signal details. At this point, increasing the window size will simultaneously smooth out the effective signal components, leading to a decrease in positioning accuracy. By calculating adjacent... The magnitude of the value change can be monitored in real time to track the variance decay rate. When this rate first falls below a preset threshold, it indicates that the system has entered the noise suppression marginal effect region. This critical point is the optimal balance between noise reduction effect and signal fidelity. By adopting a first-less-than-expected decision strategy instead of global minimum search, it is ensured that the selected window width meets performance requirements while minimizing computational overhead, achieving the optimal trade-off between efficiency and accuracy.

[0040] Preferably, in step S5, the collected backscattering Rayleigh curves are processed by moving average using the target window width to satisfy the following relationship: ; in, Indicates the use of the target window width For the original collection The first backscattering Rayleigh scattering curve obtained by moving average processing One data value, This represents the target window width. Indicates the first Backscattering Rayleigh scattering curve signal, Indicates by all The complete data sequence after processing with a moving average, arranged in sequence. and All are positive integer indices.

[0041] It should be noted that, It is a single scattering curve data generated after moving average processing; its essence is a modification of the original data. From the first curve The strips begin to flow continuously The result of averaging the curve point by point along a spatial dimension is not a single numerical value, but rather a combination of... A complete curve for each data point. Target window width. It is the optimal average number of times determined by the aforementioned variance stability determination process. This parameter remains fixed throughout step S5 and is used to perform global smoothing on all the original data. Indicates the index from arrive of The original curve is subjected to an element-wise accumulation operation, which is performed in This is performed independently at each spatial location, generating a temporary accumulation curve. Index range The starting position boundary of the sliding window is defined to ensure that the window remains within the valid data range during the sliding process. Take the maximum value At that time, the window just covered the last one. The original curve can eventually be generated. Different average curves. Complete data sequence. It is a collection The dataset of curves has a similar structure to the original data matrix, but the signal-to-noise ratio is significantly improved, providing high-quality input for subsequent moving differential localization algorithms. Positive integer index. and It is a loop variable used to access the data matrix, where Control the starting position of the sliding window. The selection of the curve involved in the averaging within the control window, together with the other two, achieves the complete moving average logic.

[0042] It is understandable that the moving average processing defined in step S5 and the parallel trial-and-error processing in step S2 are fundamentally different computational modes: step S2 optimizes parameters by only processing the preceding parameters of each candidate window width. A single average is performed on the curves to generate a small number of comparison samples; while step S5 involves adaptively determining the optimal window width. Applied to all N original curves, generated using a sliding window method. A smooth curve is used to achieve global noise reduction of the data. The physical meaning of the moving average is that it simulates the real-time processing of the traditional moving average algorithm on the time axis, with each... Corresponding to time For the follow-up The statistical average of each pulse cycle preserves the signal's temporal evolution characteristics, facilitating subsequent detection of transient disturbances through differential operations. By limiting... The range of values ​​for is defined by this relationship, ensuring that the calculation process does not access data out of bounds, thus guaranteeing the robustness of the algorithm. The generated complete sequence It includes all smoothed results from the first valid window to the last, with temporal coverage comparable to the original data. It effectively suppresses random noise while fully preserving the dynamic process of the perturbation signal, laying a data foundation for accurate positioning. This global sliding processing mode outperforms single averaging in improving the signal-to-noise ratio because it fully utilizes all possible averaging combinations, ensuring that the signal at each moment undergoes... The noise suppression effect was consistent and stable across multiple statistical averages.

[0043] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0044] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An improved moving average method applied to a distributed fiber optic vibration sensing system, characterized in that, Includes the following steps: S1: Continuously collect several backscattering Rayleigh curves, each of which consists of several data points; S2: Set several different moving average window widths, and perform moving average processing on several back Rayleigh scattering curves based on each moving average window width to obtain several moving average processed scattering curves corresponding to each moving average window width; S3: For each moving average window width, calculate the variance of the scattering curve after moving average processing at the same time point to obtain the variance sequence corresponding to the moving average window width. S4: Based on the variance sequence corresponding to each moving average window width, determine the moving average window width corresponding to when the variance change tends to stabilize and use it as the target window width; S5: Use the target window width to perform moving average processing on the collected backscattering Rayleigh curves, and locate the disturbance event on the optical fiber based on the processing result.

2. The improved moving average method for distributed fiber optic vibration sensing systems according to claim 1, characterized in that, In step S1, the data obtained from continuously collecting several backscattering Rayleigh scattering curves satisfy the following relationship: ； in, This represents a data matrix arranged as column vectors. This represents the total number of backscattering Rayleigh scattering curves collected consecutively. Indicates the first Article The backscattering Rayleigh scattering curve signal composed of 10 data points This indicates the number of data points set in a single backscattering Rayleigh scattering curve.

3. The improved moving average method for distributed fiber optic vibration sensing systems according to claim 1, characterized in that, In step S2, the moving average processing based on a given moving average window width satisfies the following relationship: ； in, Indicates the moving average window width. Indicates based on the moving average window width The first one obtained after performing moving average processing Scattering curves, Indicates the first number of the original collection The backscattering Rayleigh scattering curve signal.

4. The improved moving average method for distributed fiber optic vibration sensing systems according to claim 3, characterized in that, In step S3, the first step is to calculate the moving average obtained after performing moving average processing based on the moving average window width. The variance corresponding to the scattering curve Satisfying the relation: ； in, Indicates the moving average window width. Indicates based on the moving average window width The first one obtained after performing moving average processing Scattering curves.

5. The improved moving average method for distributed fiber optic vibration sensing systems according to claim 4, characterized in that, In step S4, the variance change is determined based on the variance sequence corresponding to each moving average window width. The variance change is obtained through the variance change sequence. To measure the variance change sequence Satisfying the relation: ； in, This indicates that when the moving average window width is The corresponding measurement value at that time Indicates the first The variance corresponding to the scattering curve The maximum value.

6. The improved moving average method for distributed fiber optic vibration sensing systems according to claim 5, characterized in that, Determining the moving average window width corresponding to when the variance change tends to stabilize and using it as the target window width includes: calculating the change amplitude between adjacent sequence values ​​in the variance change sequence, comparing the change amplitude with a preset stability threshold, and determining that the variance change tends to stabilize when the change amplitude is less than the stability threshold for the first time, and determining the moving average window width corresponding to this time as the target window width.

7. The improved moving average method for distributed fiber optic vibration sensing systems according to claim 6, characterized in that, In step S5, the collected backscattering Rayleigh curves are processed by moving average using the target window width to satisfy the following relationship: ； in, Indicates the use of the target window width For the original collection The first backscattering Rayleigh scattering curve obtained by moving average processing One data value, This represents the target window width. Indicates the first Backscattering Rayleigh scattering curve signal, Indicates by all The complete data sequence after processing with a moving average, arranged in sequence. and All are positive integer indices.

Images