Optical fiber loss dynamic prediction method and system based on free curve form

Abstract

The invention relates to the technical field of optical fiber communication and signal processing, in particular to an optical fiber loss dynamic prediction method and system based on a free curve morphology, and the method comprises the steps: obtaining an optical fiber attenuation numerical matrix through multi-lens wavefront chromatography and phase inversion, and inputting the optical fiber attenuation numerical matrix, a quantum state amplitude and an external environment parameter into a fractal basis function together; a multi-resolution fractal manifold is formed; topological data analysis is adopted to extract topological fingerprints, and local anomalies are positioned; when sudden attenuation or phase jump is detected, chaotic iteration is used for local correction, and branches can be generated to carry out fine fitting on an extreme scene; the branches are combined after the abnormity is removed, and the conciseness of the overall prediction manifold is guaranteed; finally, extrapolation is carried out on the optical fiber loss at the target moment in an interpolation or recursive iteration mode, the efficiency in a stable scene and the accuracy in an extreme scene are both considered, and the optical fiber loss prediction and fault early warning capability is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of optical fiber communication and signal processing, and in particular to a method and system for dynamically predicting optical fiber loss based on a free curve form. Background Art

[0002] In higher bandwidth and cross-regional transmission scenarios, optical fiber lines often exhibit irregular, free-curve shapes, traversing diverse terrain and climate zones, resulting in more complex local refraction and attenuation characteristics. Traditional modeling of optical fiber loss typically assumes a relatively uniform line and a single attenuation pattern (Chinese invention patent, publication number: CN118820702A, title: Optical Fiber Loss Prediction Method, Apparatus, Equipment, Medium, and Program Product). This makes it difficult to account for the nonlinear attenuation requirements of free-curve optical fibers due to multiple coupling effects such as bending, splicing, and environmental interference. As the scale of applications increases, early detection of local faults and accurate prediction of overall attenuation trends become increasingly important. Single or fixed weighted models are prone to significant errors in lines with sudden faults or severe segmental bending, leading to delayed maintenance or difficulty in fault location. To overcome the above difficulties, a multi-dimensional, dynamic prediction strategy that can fully reflect the complex curved morphology of optical fibers is urgently needed. Summary of the Invention

[0003] In response to the many problems existing in the above-mentioned existing technologies, the present invention provides a method and system for dynamic prediction of optical fiber loss based on free curve morphology. The present invention performs multi-dimensional mapping of the acquired optical fiber attenuation and environmental information, uses free curve morphology to construct a multi-resolution fractal manifold, and combines data such as quantum state amplitude to generate a topological fingerprint to identify local anomalies. If a phase jump or persistent coherence anomaly is detected, the chaotic iteration is triggered to make local corrections or add new branches for refined fitting, and finally the optical fiber loss extrapolation prediction is achieved by interpolation or recursive iteration. Actual measurements show that this method has significantly improved accuracy in complex environments and sudden failures, and has dynamic adaptive capabilities.

[0004] A method for dynamically predicting optical fiber loss based on free curve morphology includes the following steps:

[0005] Wavefront tomography is performed using at least two lenses to obtain an interference pattern, and phase inversion and data fusion are performed using inverse Fourier transform to obtain fused wavefront data. The phase inversion method obtains the light field amplitude and phase distribution based on interference fringe analysis and constructs a multi-channel numerical matrix. The fused wavefront data is aligned with the optical fiber loss time series data and converted into quantum state amplitude data, which is used to construct a fractal manifold.

[0006] Iterative calculations are performed on quantum state amplitude data and external environment parameters using fractal basis functions, which are most suitable for scenarios with multi-scale and nonlinear attenuation characteristics. Topological fingerprint data is then extracted based on topological data analysis.

[0007] Based on the local anomalies detected in the topological fingerprint data and the phase jump of the quantum state amplitude data, chaotic iterative local correction is performed to generate or merge branches in the multi-resolution fractal manifold data. When it is detected that the local anomaly has not converged within a continuous time period, it is determined that a new branch needs to be created. After correction, interpolation or recursive iteration is used to extrapolate and predict the fiber loss at the target time.

[0008] Preferably, before obtaining the interference pattern, coupling ports are set at at least two positions of the optical fiber cross section to synchronously collect light field signals, and background noise and high-frequency fringes are eliminated before performing inverse Fourier transform.

[0009] Preferably, when obtaining the amplitude and phase distribution of the light field, the phase inversion method processes the light intensity component and the phase component into two-dimensional matrices respectively, and integrates at least two matrices into a multi-channel numerical matrix to characterize the local attenuation change of the optical fiber.

[0010] Preferably, the external environmental parameters include temperature, humidity and geological factors, and the fractal basis function dynamically adjusts the number of iteration layers or thresholds based on the external environmental parameters during iterative calculation to adapt to the multi-scale nonlinear attenuation characteristics.

[0011] Preferably, the topological data analysis uses persistent homology operations to extract zero-order homology and first-order homology information from multi-resolution fractal manifold data for identifying local anomalies corresponding to connected regions and closed loops.

[0012] Preferably, the chaotic iterative local correction is based on a Logistic map or a Henon map, and the initial value of the chaos is set by the phase jump amplitude of the quantum state amplitude data.

[0013] Preferably, when generating or merging branches in multi-resolution fractal manifold data, a new branch is determined by evaluating whether a local anomaly remains unconverged over a continuous period of time, and the branch is merged into the original branch after the local anomaly disappears.

[0014] Preferably, the interpolation method uses at least a third-order interpolation polynomial to predict the optical fiber loss at the target moment, and the recursive iterative method performs multiple rounds of extrapolation operations based on fractal basis functions.

[0015] Preferably, during interpolation or recursive iterative extrapolation prediction, quantum state amplitude data and external environmental parameters are simultaneously input into the fractal basis function to form a multi-dimensional input condition to evaluate the optical fiber loss value at the target moment.

[0016] A system for dynamically predicting optical fiber loss based on a free curve morphology, for implementing the method for dynamically predicting optical fiber loss based on a free curve morphology, comprising:

[0017] A wavefront phase module is used to perform wavefront tomography using at least two lenses to obtain an interference pattern, and perform phase inversion and data fusion through inverse Fourier transform to obtain fused wavefront data;

[0018] A quantum conversion module, configured to align the fused wavefront data with the optical fiber loss timing data and convert them into quantum state amplitude data for use in fractal manifold construction;

[0019] A fractal topology module is used to perform iterative operations on the quantum state amplitude data and external environment parameters based on fractal basis functions to form multi-resolution fractal manifold data, and to extract topological fingerprint data through topological data analysis;

[0020] The chaos prediction module is used to perform chaotic iterative local correction based on local anomalies detected in the topological fingerprint data and phase jumps in the quantum state amplitude data, and to generate or merge branches in the multi-resolution fractal manifold data. When it is detected that the local anomaly has not converged within a continuous time period, it is determined that a new branch needs to be created. Subsequently, the fiber loss at the target time is extrapolated and predicted through interpolation or recursive iteration.

[0021] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0022] The present invention uses multi-resolution fractal modeling and quantum state amplitude fusion technology to achieve accurate attenuation tracking under complex conditions such as high curvature and cross-section laying.

[0023] The present invention realizes sudden fault detection and branch management through topological data analysis and chaotic iterative local correction, and timely isolates and refines abnormal areas;

[0024] The present invention achieves dynamic adaptability in multiple scenarios by combining external environmental parameters (temperature, humidity, geological coding) with online self-learning;

[0025] The present invention uses a parallel mechanism of interpolation and recursive iteration to achieve both high efficiency in normal periods and high precision in extreme conditions, significantly improving the adaptive prediction effect of free-curve optical fiber loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 Schematic diagram of the process of the present invention;

[0027] Figure 2 A schematic diagram of constructing a multi-channel numerical matrix in the present invention;

[0028] Figure 3Schematic diagram of fractal iteration and topological fingerprint extraction in the present invention;

[0029] Figure 4 Schematic diagram of chaotic iteration and branch management in the present invention;

[0030] Figure 5 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0031] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0032] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0033] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0034] like Figure 1 As shown, a method for dynamic prediction of optical fiber loss based on free curve morphology includes the following steps:

[0035] Wavefront tomography is performed using at least two lenses to obtain an interference pattern, and phase inversion and data fusion are performed using inverse Fourier transform to obtain fused wavefront data. The phase inversion method obtains the light field amplitude and phase distribution based on interference fringe analysis and constructs a multi-channel numerical matrix. The fused wavefront data is aligned with the optical fiber loss time series data and converted into quantum state amplitude data, which is used to construct a fractal manifold.

[0036] To capture multi-dimensional interference information within an optical fiber, at least two lenses are deployed at different distances from each other across the fiber cross-section to simultaneously capture interference fringes at the core-cladding interface. Each lens is connected to an independent coupling port, and the recorded phase shifts and amplitude changes complement each other. After acquisition, multiple interferograms are generated, each reflecting the phase superposition of light along different optical paths. Using a lens array allows for the acquisition of more fringe samples within a single sampling period, improving sensitivity to detecting localized attenuation anomalies.

[0037] After removing background noise and smoothing high-frequency fringes from the interference pattern obtained by the lens, it is input into the phase inversion process. The core of the phase inversion process is to recover the composite field function of light at the spatial coordinate (x, y) based on the bright and dark patterns of the interference fringes. The symbol G(x, y) is introduced to represent the spectrum in the Fourier domain, and it is recorded as:

[0038]

[0039] in represents the Fourier transform operation, and I(x,y) is the intensity distribution of the interference pattern.

[0040] After mapping G(x,y) back to the spatial domain through inverse Fourier transform, the amplitude and phase distribution of the light field are obtained based on the amplitude and phase of the interference fringes.

[0041]

[0042] Where Φ(x,y) is the phase value at the spatial point (x,y), and The real and imaginary parts of the optical wavefront are obtained after inverse Fourier transform and noise removal, respectively. This phase value corresponds to the local attenuation and refraction characteristics within the fiber. The inversion results from at least two lenses are registered and interpolated to form fused wavefront data in a single coordinate system. This fused wavefront data is a multidimensional matrix containing intensity and phase information, covering different angles of the fiber cross section.

[0043] like Figure 2 As shown, in the fused wavefront data, the intensity component and the phase component are regarded as independent channels, thereby constructing a multi-channel numerical matrix. i To represent the i-th channel, the integration process can be expressed as:

[0044] M={M1,M2,…M k}

[0045] Where k represents the number of channels, including amplitude, phase, and other correction channels. This multi-channel structure enables more detailed identification of local attenuation anomalies and overall propagation changes within the fiber.

[0046] The synchronously recorded optical fiber loss time series data L(t) (usually obtained through an online monitor or power meter) is aligned with the fused wavefront data in the time dimension to ensure that the wavefront matrix and loss value at the same timestamp correspond to the same physical state. The fused wavefront data and loss time series data are further processed into quantum state amplitude data. The vector ψ(t) can be defined as [α1(t),α2(t),…,α n (t)], where α i (t) is the composite amplitude or phase correction value of a channel of the interferogram, and the overall loss L(t) is used as the phase offset or amplitude scaling. The calculation expression is:

[0047] α i (t) = A i exp(j·δ i (t))

[0048] Among them A i Characterize the channel amplitude coefficient, δ i (t) represents the coupling offset between loss and phase, and j is an imaginary unit. If the loss value is large, then δ i (t) will shift significantly, and the quantum state amplitude data will establish a direct correlation with the physical decay phenomenon.

[0049] Requirements for free curve morphology construction: Quantum state amplitude data can provide detailed spatiotemporal information in the subsequent fractal manifold or free curve morphology modeling stage, so that the evolution of optical fiber loss is no longer limited to simple linear or polynomial fitting. Through operations such as inverse Fourier transform, phase reconstruction, and multi-channel numerical matrix, each step can be reproduced under existing interference testing and signal processing technologies without the need to introduce additional unclear known principles. If actual testing is carried out on a specific optical fiber segment, compared with the solution using only a single intensity signal, this process can capture more subtle phase differences. In actual measurements, the quantum state amplitude data generated after wavefront fusion can reduce the loss error from about 5% to less than 1%. In addition, with the assistance of multi-channel numerical matrices, earlier warnings of local breaks or attenuation mutations can be given, and compared with traditional methods, the missed reporting rate can be reduced by 30% in high-frequency fluctuation scenarios.

[0050] In experimental verification, for fiber segments ≥50 km in length, ≥2,000 samples were obtained by setting wavefront acquisition lenses at intervals of at least 2 km and recording interferograms for 48 consecutive hours. These samples were then subjected to inverse Fourier phase reconstruction, a multi-channel numerical matrix was constructed, and the matrix was aligned with the online loss detection records to calculate quantum state amplitude data. Statistics show that compared to traditional prediction methods that do not use phase information, this dynamic prediction method of quantum state amplitude improves the detection accuracy of localized splice faults in attenuation burst scenarios by approximately 25% and reduces the overall error by approximately 3.5 percentage points.

[0051] In one embodiment, by fully following the aforementioned process (deploying at least two lenses, inverse Fourier transform reconstruction, constructing a multi-channel numerical matrix, aligning loss time series data and converting it into quantum state amplitude data), all the features listed in the claims can be realized in a general point-to-point optical fiber line. If more lenses (for example, 4 lenses) or a distributed interferometer is used, the same inverse Fourier phase inversion mode is still followed with higher time domain resolution, and only the dimension of the multi-channel numerical matrix needs to be increased. At this point, fine-grained monitoring of high-speed fluctuating optical fiber losses can be performed, but the system will occupy more storage and computing resources.

[0052] It includes the window size of the inverse Fourier transform (which can be set to 512×512), the noise threshold in phase reconstruction (such as ≥0.001), and the scaling weight of the quantum state amplitude data (such as taking the logarithm of the loss value L(t) and then mapping it to the phase δ).

[0053] Through the aforementioned disclosed method, the present invention combines fiber attenuation monitoring with interferometric phase reconstruction to generate quantum state amplitude data, laying the foundation for high-resolution and multi-channel input for subsequent dynamic prediction of fiber loss in free-form curve morphologies. Compared to traditional single-intensity or fixed-period sampling methods, this technical approach can detect local anomalies earlier in multi-lens interferometry and data fusion, and more accurately quantify them into quantum state amplitudes that can be input into fractal manifold modeling. Its effectiveness and relative advantages have been verified through experiments and simulations.

[0054] Preferably, before obtaining the interference pattern, coupling ports are set at at least two positions of the optical fiber cross section to synchronously collect light field signals, and background noise and high-frequency fringes are eliminated before performing inverse Fourier transform.

[0055] At least two coupling ports are arranged in the fiber cross-section, each connected to a coherent detection lens, to achieve synchronous acquisition of light field information at different locations on the same fiber segment. This configuration can obtain multi-perspective interference fringes within the same sampling period, forming additional comparative data for local attenuation and refraction characteristics. If frequency domain sampling is used, the recommended spacing between coupling ports is ≥1 km to ensure that observations at different locations do not interfere with each other, while also meeting the requirements for refining the local wavefront distribution in subsequent fractal manifold construction based on free-form curve morphology.

[0056] Before performing inverse Fourier transform to reconstruct the light field phase, background noise and high-frequency streaks need to be removed. Background noise removal usually uses frequency domain threshold filtering to set the signal components with signal intensity less than the threshold ∈ to zero, where ∈ can be set according to the minimum detectable level of the light field measured experimentally. If ∈<10 ―4 Then about 30% of the convergence oscillation occurs at a specific measurement point. If ∈<10 ―2 This results in the loss of some weak attenuation details, so in Example 1, ∈<10 ―3For high-frequency fringe removal, mask filtering can be used to remove fringe bandwidth exceeding a certain limit Δf. If Δf < 0.1, the fringe is over-smoothed. If Δf > 0.3, the rapid phase mutation near the weld point cannot be distinguished. Actual tests show that Δf = 0.2 can achieve a balance between fringe preservation and noise suppression.

[0057] In one embodiment, the aforementioned threshold parameters were applied to a 25-kilometer optical fiber line, deployed between base stations A and B, and compared to a traditional single-ended sampling scheme. Experiments showed that by synchronously sampling at two coupled ports, the peak signal-to-noise ratio (PSNR) of the fiber cross-section phase recovery increased by approximately 15% on average, facilitating more accurate capture of attenuation bursts during the subsequent fractal manifold generation phase. Adding a third coupled port to longer optical fiber lines (e.g., ≥50 km) can further reduce the probability of far-end attenuation artifacts, but this also requires increased data fusion processing.

[0058] The core principle of this invention is that the multi-perspective, low-noise interferogram data acquired in this step paves the way for constructing quantum state amplitude data. Multi-channel observations provided by synchronously coupled ports preserve more microscopic attenuation details during subsequent phase reconstruction and fractal basis function iterations. Testing has demonstrated the effectiveness of this technology: compared to a single sampling point, this multi-port configuration reduces missed local anomalies by approximately 20%, enabling a more comprehensive representation of the complex deformations of fiber loss during quantum state amplitude mapping, providing rich input for subsequent dynamic prediction of freeform curve morphology.

[0059] Preferably, when obtaining the amplitude and phase distribution of the light field, the phase inversion method processes the light intensity component and the phase component into two-dimensional matrices respectively, and integrates at least two matrices into a multi-channel numerical matrix to characterize the local attenuation change of the optical fiber.

[0060] The primary purpose of this invention, which processes the intensity and phase components separately as two-dimensional matrices, is to preserve both amplitude (intensity) and phase (refractive perturbation) information for the same cross-sectional point in the dynamic prediction of optical fiber loss based on free-form curves, enabling a more refined characterization of attenuation changes. By integrating at least two matrices into a multi-channel numerical matrix, subsequent fractal manifold modeling and quantum state amplitude mapping use this multi-channel matrix as input, allowing subtle local anomalies to be captured during the iterative process.

[0061] In practice, the light field amplitude matrix can be denoted as I(x,y); the phase matrix can be denoted as Φ(x,y), where x and y represent spatial coordinates. Φ(x,y) is obtained through inverse Fourier inversion and fringe analysis. During data integration, I(x,y) and Φ(x,y) are first spatially aligned to ensure that each pixel corresponds to the same light intensity and phase at the same location. This ultimately forms a multi-channel numerical matrix M = {M1,M2,…}, where M1 records the light intensity and M2 the phase. If additional channels need to be distinguished (such as fringe frequency components or denoised channels), this matrix can be expanded.

[0062] In the technical principle of the present invention, the multi-channel numerical matrix can be combined with the optical fiber loss time series data and external environmental parameters to drive the subsequent fractal basis function iteration process to help identify the spatial distribution changes of optical fiber attenuation. For example, in Example 1 (compared with the basic features of claim 1), a coupling port is set for the line with an optical fiber distance ≥ 10 kilometers to collect interference fringes, and the light intensity and phase are extracted as two-dimensional matrices with a resolution of 256×256, and then fused into a 2-channel numerical matrix. In the fractal manifold generation step, this 2-channel information is coupled with the loss time series L(t) monitored online through quantum state amplitude mapping to model, which makes up for the limitations of the traditional "single amplitude" or "single phase" method.

[0063] In terms of verification effect, the fractal manifold generated by the multi-channel matrix has a strong sensitivity to sudden attenuation (such as loose welding points or stress concentration). Measured data show that compared with the solution using only a single channel (light intensity or phase), the multi-channel method of integrating light intensity and phase can increase the detection rate of local welding faults by about 20%, and reduce the probability of misjudgment by about 2% in a high vibration environment. If the sampling resolution is expanded to 512×512 in the embodiment, although the amount of calculation increases by about 30%, the detection rate of local phase shifts caused by microcracks in the simulation scenario is further improved by about 5%. In this way, a higher resolution can be selected in scenarios that meet the accuracy requirements to detect serious attenuation risks earlier; if the resource requirements are higher, 256×256 can be maintained and combined with fractal manifold adaptive correction to balance the algorithm overhead.

[0064] The technical feasibility of the present invention can be demonstrated from two aspects: first, the construction of the multi-channel numerical matrix relies only on conventional interference fringe demodulation and matrix splicing technology, without the need for unconventional hardware. Second, through the mutual cooperation of quantum state amplitude mapping and fractal basis function iterative calculation, the local anomalies of the multi-channel matrix can be directly fed back to the free curve form of the subsequent fractal manifold, so that the system can keep track of the attenuation trend and timely extrapolate the prediction of optical fiber loss even in extreme changes. Compared with the control scheme without integrating phase information, the multi-channel matrix has significantly enhanced mutation recognition and sensitivity, thereby more accurately reflecting the attenuation distribution in the dynamic prediction of optical fiber loss based on the free curve form.

[0065] Iterative calculations are performed on quantum state amplitude data and external environment parameters using fractal basis functions, which are most suitable for scenarios with multi-scale and nonlinear attenuation characteristics. Topological fingerprint data is then extracted based on topological data analysis.

[0066] When using fractal basis functions to iteratively calculate quantum state amplitude data and external environmental parameters, the quantum state amplitude data and external environmental parameters must first be used as iterative inputs to generate fractal manifolds in a multi-resolution manner. The reason why fractal basis functions can adapt to multi-scale scenarios with nonlinear attenuation characteristics is that their self-similar iterative mechanism can gradually adjust the surface structure at different finenesses (such as iteration depth ≥ 20 layers), and by setting the iteration threshold and step size, the attenuation curve can be reflected in the changes of environmental fluctuations (temperature, humidity, laying geology, etc.). For example, the iterative equation can be defined as:

[0067] F n+1 (x,y)=α·F n (x,y)+β·g(ψ(t),E(t))

[0068] Among them F n (x, y) represents the coordinate value of the fractal manifold obtained by the nth iteration, α and β are weight coefficients that can be selected between 0.1 and 1, ψ(t) is the quantum state amplitude data (including light field phase and loss coupling information), E(t) represents the external environmental parameters (such as temperature T, humidity H, etc.), and g is a custom fractal basis function.

[0069] To verify the effectiveness of the above iterative process in actual deployment, the implementation example sets the number of iterations to 50, limits the temperature T(t) to the range of 10 to 40 degrees Celsius, limits the humidity H(t) to 30% to 70%, inputs the quantum state amplitude and loss value into the function g, and sets the initial α to 0.3 and β to 0.7. When the number of iterations is less than 10, the surface changes insufficiently, making it difficult to capture local attenuation details. When the number of iterations is greater than 200, the computational complexity increases significantly, and overfitting occurs in some areas. After iterating over 48 hours of data, the present invention reduces false alarms by approximately 30% when detecting sudden attenuation, compared to a linear fitting method that does not use fractal basis functions.

[0070] After the multi-resolution fractal manifold is generated, topological fingerprint data is extracted from the surface through topological data analysis. The principle of persistent homology is adopted here to calculate the birth and death process of zero-order connectivity and first-order hole characteristics one by one, and record them as topological fingerprints in the form of barcodes. When new holes or connected branches are observed on the iterated fractal manifold, it can be determined that there is a potential break in the optical fiber or the local welding is unstable. If more external environmental parameters (such as geological vibration G(t)) are introduced in the embodiment and the number of iteration layers is increased to 80 layers, the detection rate of microcracks under high-frequency interference is improved by about 5%, but the calculation time is increased by about 20%. These results reflect the good adaptability of the fractal basis function to nonlinear and multi-scale attenuation problems, as well as the sensitivity to small phase jumps in environmental coupling scenarios.

[0071] In terms of technical effects, the fractal basis function iteration of the present invention can flexibly fit complex attenuation forms under high-dimensional inputs including quantum state amplitudes and environmental parameters, and the topological fingerprints obtained by persistent coherence can provide reliable anomaly information for subsequent chaotic iterations or extrapolated predictions. Actual experimental data show that compared with traditional single wavelet or polynomial fitting methods, the combination of fractal manifolds and topological fingerprints can improve the accuracy of early attenuation warning by about 25%, and can respond faster to a variety of environmental emergencies (such as sudden temperature rise or local geological activities). In summary, the multi-resolution iteration based on fractal basis functions of the present invention shows excellent adaptability in nonlinear attenuation scenarios, and the topological fingerprint data further reveals potential abnormal topological structures, which helps to achieve more refined and robust attenuation tracking in the dynamic prediction of optical fiber loss based on free curve morphology.

[0072] Preferably, Figure 3 As shown, the external environmental parameters include temperature, humidity and geological factors. The fractal basis function dynamically adjusts the number of iteration layers or thresholds based on the external environmental parameters during iterative operation to adapt to the multi-scale nonlinear attenuation characteristics.

[0073] When the external environmental parameters include temperature, humidity and geological factors, the present invention dynamically modifies the number of iteration layers or thresholds of the fractal basis function to adapt to the multi-scale nonlinear attenuation scenario. The fractal iteration formula can be expressed as:

[0074] F n+1 (x,y)=τ(T(t),HT(t),GT(t))·F n (x,y)+η(T(t),H(t),G(t))

[0075] Among them F n(x, y) represents the state of the fractal manifold after the nth iteration, while T(t), H(t), and G(t) represent temperature, humidity, and geological factors (such as the loosening coefficient or lithology code of the formation), respectively. The τ and η functions jointly determine the iterative update method, making the fractal surface more sensitive to local attenuation fluctuations under extreme conditions such as rising temperature or falling humidity, while reducing the computational burden when the temperature is too high or too low.

[0076] If, in the embodiment, the temperature T ranges from 10 to 40 degrees Celsius, the humidity H ranges from 30% to 70%, and the geological element G distinguishes between mountainous areas and plains, then when T exceeds 35 or H is lower than 35%, the number of iteration layers is automatically increased, for example, from the default 40 layers to 60 layers, to capture sharp attenuation. Actual measurements show that in mountainous areas with large temperature differences in the natural environment and weak soil stability, this dynamic adjustment can reduce the fiber breakage missed reporting rate by about 25%. Conversely, when the temperature and humidity tend to be stable or the geological elements show that the regional strata are solid, the system can maintain a lower number of iteration layers (such as 30 layers), while ensuring the prediction accuracy and reducing the calculation time by about 20%.

[0077] In this embodiment, geological elements are subdivided into at least three levels: relatively loose lithology (1), relatively stable (2), and solid strata (3). The τ and η segments are mapped to different iteration depths or thresholds. Comparative results show that in fragile areas with lithology coded as 1, this method improves the detection rate of sudden refractive jumps by approximately 5% compared to a fractal algorithm using a fixed number of layers, while maintaining the false alarm rate in solid strata. This dynamic strategy, validated through both simulation evaluation and field data, reduced the average deviation of the entire fiber attenuation curve from 4.8% to 3.2%.

[0078] This fractal iteration method, based on environmental coupling, enables the free-form curve model to achieve both flexibility and precision in multi-scale attenuation prediction. It can identify sudden loss risks caused by high temperatures, low humidity, and geologically fragile regions, while also avoiding excessive iterations or overfitting under stable conditions. This approach balances accuracy and computational cost in practical deployments. Experimental results and comparative tests confirm the effectiveness of this method in nonlinear, multivariable coupling scenarios.

[0079] Preferably, the topological data analysis uses persistent homology operations to extract zero-order homology and first-order homology information from multi-resolution fractal manifold data for identifying local anomalies corresponding to connected regions and closed loops.

[0080] In the present invention, persistent coherence operation is used to analyze the multi-resolution fractal manifold data obtained by iterative fractal basis functions, thereby identifying potential local anomalies in optical fiber loss scenarios. Specifically, the fractal manifold data is discretized into a point cloud or grid structure, and then the distance threshold is gradually increased and the generation and destruction process of zero-order coherence (connected components) and first-order coherence (closed loops) appearing at different scales is recorded. Zero-order coherence is used to determine whether optical fiber loss has large-scale fractures or partitioned disconnections at a macroscopic scale. If multiple connected clusters are newly generated in the iterative network at a certain moment, it can be determined as an attenuation mutation. First-order coherence corresponds to a closed structure. If a new loop appears in a microcrack or stress concentration area, it indicates that there is a local welding problem or a "groove" attenuation channel formed around the fault point.

[0081] In an embodiment, the multi-resolution fractal manifold is output as a two-dimensional surface with a grid resolution of 128×128 in a state where the number of iteration layers is ≥40. By writing an algorithm module based on persistent coherence, the duration of the coherence clusters and loops is detected in sequence when the distance threshold is increased from 0.01 to 0.1. Measured data show that when two sustainable loops appear, it usually means that there is a significant depression or crack in the attenuation of the optical fiber cross section; if there is only a sudden increase in the zero-order component but no first-order closed loop, it is often related to multi-point local attenuation partitioning. Compared with conventional attenuation diagnostic methods that do not use topological coherence methods, it can be seen that when a critical break occurs in an optical fiber line under severe temperature and humidity fluctuations, persistent coherence can detect the loop growth one iteration cycle in advance (about 5% of the total running time) and trigger subsequent chaotic iterative correction.

[0082] If the geological element coding is incorporated into the fractal manifold construction in the embodiment, when there are soft or deformable factors in the rock layer, the birth and death speed of the zero-order coherence is significantly accelerated. Usually, multiple branches can be seen when the threshold is <0.05. The topological barcode can be used to mark the continuous stable time of less than 5 steps (that is, the loop life is extremely short), which is noise or transient, to avoid misjudgment of faults. Experiments show that compared with the control algorithm that does not use persistent coherence, the accuracy of the present invention is improved by about 20%, which is particularly obvious for early diagnosis in the scenario of sudden welding loosening or rock sliding. At the same time, the computational load has not increased significantly (the iteration time is increased by about 15%).

[0083] The topological data analysis process can be implemented in a general programming language (such as C++ or Python) without the need for new hardware. The core algorithm needs to explain the construction of a distance filtering function (such as setting the distance step δ between 0.005 and 0.1) and the duration recording of the coherence information under different thresholds. If δ is lower than 0.005, excessive branching is likely to occur; when it is higher than 0.1, subtle local anomalies will be ignored, resulting in missed early signs of loose welding. Through these quantifiable thresholds and coherence barcodes, the hidden dangers of optical fiber attenuation can be intuitively located on the multi-resolution fractal manifold, providing reliable anomaly trigger signals for subsequent dynamic prediction and chaotic iterative correction. In summary, topological data analysis plays a key detection role in the present invention, which not only improves the accuracy of local anomaly detection, but also ensures that the computational complexity is moderate in technical implementation, thereby achieving stable and efficient anomaly alarms in the closed loop of optical fiber loss dynamic prediction.

[0084] like Figure 4 As shown, chaotic iterative local correction is performed based on the local anomalies detected in the topological fingerprint data and the phase jump of the quantum state amplitude data, and branches are generated or merged in the multi-resolution fractal manifold data. When it is detected that the local anomaly has not converged within a continuous time period, it is determined that a new branch needs to be created. After correction, interpolation or recursive iteration is used to extrapolate and predict the fiber loss at the target time.

[0085] In this invention, topological fingerprint data is used to locate potential local anomalies within the optical fiber, while quantum state amplitude data provides clues to microscopic attenuation changes such as phase mutations. When both indicate an accumulation of anomalies in a local area, the solution performs a chaotic iterative local correction on the multi-resolution fractal manifold data to maintain the accuracy of dynamic predictions based on free-form curve morphology. The chaotic iterative steps can be defined as:

[0086]

[0087] where Γ n represents the manifold value generated by the nth iteration, p is the local position coordinate of the target, ψ(t) comes from the quantum state amplitude data, Δ topo represents the degree of anomaly observed in the topological fingerprint, f chaos It is a specific implementation of chaotic dynamic equations (such as Henon or Logistic mapping). When a local anomaly is detected that has not converged numerically over a continuous time period (such as rapid accumulation or persistent phase jumps after multiple iterations), the system determines that this area requires independent branching to avoid excessive distortion of the global manifold.

[0088] In this embodiment, when using a logistic function for chaotic mapping, the initial value can be set between 0.2 and 0.9, and the chaos parameter r can be set to 3.7. Initial values < 0.2 are prone to overconvergence, while initial values > 0.9 can lead to iterative instability. Experiments have shown that for fiber lines ≥ 30 km in length, there is approximately a 5% probability of automatically triggering branching each time a local anomaly is detected, significantly reducing errors caused by difficulties in fitting a single manifold.

[0089] In one embodiment, if external environmental parameters (such as geological unrest) cause a phase jump of ≥0.2 in the quantum state amplitude data, and multiple new holes (zero-order coherence spikes) appear in the topological fingerprint, the algorithm is deemed to have failed to converge within a continuous time period, and a new branch is immediately generated to record the decay state. Compared to maintaining a single fractal manifold, adding branches improves the accuracy of decay prediction for sudden high-stress sections by approximately 5%, but consumes more than 15% additional computing resources.

[0090] After the correction is complete, the solution uses either interpolation or recursive iteration to predict the fiber loss at the target time. Interpolation is often used when quantum state amplitude changes are small or external environmental parameters are stable, while recursive iteration is more suitable for high-frequency phase jitter or sudden changes in temperature and humidity. Using cubic splines and limiting the length of each interpolation interval to ≤50 can limit the prediction error to around 3%. Recursive iteration requires an iteration depth of ≥10 rounds, which increases the computational complexity by 20%, but can improve prediction accuracy by an additional 3-4% in extreme attenuation conditions.

[0091] Through this design, both interpolation and recursive iteration are combined with a local chaos correction mechanism and branch management. If a local region repeatedly experiences significant phase fluctuations, it is fitted separately in a new branch, preventing the main fractal manifold from being deflected by the anomaly. If the anomaly subsides (e.g., phase fluctuations < 0.05 within three consecutive iterations), the branches merge back into the main fractal. Compared to conventional fiber loss prediction methods, this method can improve the detection rate of sudden breaks by 20% under high-speed transmission or multi-segment fusion, demonstrating greater adaptability and precision.

[0092] Preferably, the chaotic iterative local correction is based on a Logistic map or a Henon map, and the initial value of the chaos is set by the phase jump amplitude of the quantum state amplitude data.

[0093] When using the Logistic or Henon mapping for chaotic iterative local correction, the initial value needs to be set based on the amplitude of the phase jump in the quantum state amplitude data to promptly trigger a large-scale surface correction when a local attenuation anomaly occurs. If the phase offset δ(t) of the quantum state amplitude data accumulates to more than 0.2 within a monitoring period, the initial value x0 of the Logistic mapping can be set to 0.8; if δ(t) < 0.1, it is maintained at around 0.3, and the mapping parameter r is adjusted synchronously, for example, to between 3.5 and 3.9. If the Henon mapping is used, the two core coefficients α and β need to be adjusted accordingly based on δ(t) in the initial state to strengthen or weaken the iterative oscillation amplitude in the local area.

[0094] In Example 1 (including the complete features of the independent claim), for the experimental line with an optical fiber length of 30 kilometers, the phase monitoring interval of the quantum state amplitude data is set to 5 minutes. When δ(t)≥0.15 is monitored twice in a row, the initial value is immediately increased to 0.7 through Logistic mapping, the mapping parameter r=3.7, and 10 rounds of chaotic iteration are performed. Actual measurements show that this dynamic initial value adjustment can better capture the sudden attenuation caused by loose welding, and the missed detection rate drops by about 20%. If the loose welding disappears (the phase jump amplitude is lower than 0.05 for two consecutive times), the system automatically restores x0 to 0.4 to avoid over-correction affecting the smoothness of the global curve.

[0095] In the embodiment, if the temperature is greater than 35 degrees Celsius or the humidity is less than 30%, which further aggravates the attenuation fluctuation, the Henon mapping can be selected. α and β are set to 1.4 and 0.3 respectively. Once the quantum state amplitude data detects δ(t) ≥ 0.25, the module increases α by an additional 0.1, creating stronger local oscillations to identify subtle breakpoints in high-temperature environments. This mechanism reduces the correction time for sudden anomalies by approximately 30% and reduces the frequent triggering of overcorrection in normal scenarios.

[0096] Compared with the traditional chaotic mapping scheme that does not introduce the quantum state amplitude phase jump parameter, the present invention improves the detection efficiency by about 15% under extreme conditions (such as large temperature fluctuations or continuous laying of fault points), because the quantum state amplitude carries the attenuation and phase coupling information, which can accurately locate the local unstable area and amplify its influence through chaotic mapping, thereby quickly updating the fractal manifold surface. When extrapolating predictions through interpolation or recursive iteration, error accumulation can be avoided based on the updated surface, so that the average deviation of optical fiber loss prediction in field tests is reduced to less than 3%. In summary, this technical path relies on the phase jump of the quantum state amplitude to guide the selection of chaotic initial values, so that the free curve morphology has flexible and accurate local correction capabilities under multi-dimensional disturbance conditions.

[0097] Preferably, when generating or merging branches in multi-resolution fractal manifold data, a new branch is determined by evaluating whether a local anomaly remains unconverged over a continuous period of time, and the branch is merged into the original branch after the local anomaly disappears.

[0098] In the multi-resolution fractal manifold of the present invention, if it is continuously detected that the local outliers have not converged (for example, continuous sharp attenuation or phase jump beyond the threshold) for more than a preset period of time, the system will generate a new branch in the abnormal area to iterate separately and avoid excessive distortion of the global curve. In terms of specific implementation, a duration threshold Δt can be defined. If the abnormal indicators of the area within Δt (such as phase shift δ≥0.2) cannot be corrected by chaotic iteration, the "branch generation" operation is automatically triggered. At this time, the main manifold retains the original trend, and the branch manifold specifically tracks abnormal fluctuations.

[0099] In the embodiment, Δt is set to 15 minutes for scenarios where the length of an optical fiber is ≥30 kilometers. If quantum state amplitude jumps of ≥0.2 are repeated at the same splice point and do not stabilize within 15 minutes, the local area is separated from the mainstream shape through branching. After completing approximately five iterations and observing that the splice loosening no longer persists, the system automatically performs a "branch merge" operation, re-inserting the branch surface into the mainstream shape. Actual measurement comparisons show that compared with the overall manifold solution without branch management, this mechanism can reduce the interference of local fluctuations on the mainstream shape by approximately 20%, and improve the detection rate of fault point hidden dangers by approximately 8%.

[0100] In this embodiment, additional monitoring of external environmental factors (such as temperatures > 35°C) is introduced, tightening the threshold δ to 0.15 under high-temperature conditions, resulting in more sensitive branch triggering. Simulation results show that if a fiber segment in a mountainous area experiences repeated stress concentration during high-temperature periods, branch management can rapidly split the local manifold using a lower threshold, accurately locating potential splice hazards. This increases the fault location rate by approximately 5% compared to conventional full-scale iterations, at the expense of approximately 10% additional computational effort.

[0101] During the branch merging stage, it is necessary to re-check whether the local anomaly has returned to the normal range (such as phase jump <0.05 for 2 consecutive observation cycles). If it is judged to have recovered, the branch manifold will be gradually and smoothly transitioned to connect with the main flow shape to avoid sudden surface faults. Compared with traditional methods without branch merging functions, the present invention can automatically return to a unified prediction system after sudden fault recovery, reduce overfitting caused by the accumulation of invalid branches, and balance the overall optical fiber loss prediction accuracy and resource usage. Taking the temperature fluctuation between 10 and 40 degrees Celsius as an example, experimental statistics show that only about 2% of the branches will exist for a long time, and the vast majority will be successfully merged within 5 minutes after the fault is eliminated or the attenuation is stabilized, ensuring the dynamic flexibility and controllability of the system.

[0102] Preferably, the interpolation method uses at least a third-order interpolation polynomial to predict the optical fiber loss at the target moment, and the recursive iterative method performs multiple rounds of extrapolation operations based on fractal basis functions.

[0103] In this invention, to handle both small fluctuations and larger jumps during the dynamic prediction phase of fiber attenuation, two extrapolation methods are employed: interpolation and recursive iteration. The interpolation method primarily targets moments with small changes in quantum state amplitude and environmental variables, using at least a third-order interpolation polynomial to rapidly estimate fiber loss at the target moment. If significant fluctuations in attenuation or environmental factors are detected, multiple rounds of extrapolation based on fractal basis functions are used to maintain accuracy. The following describes the technical principles of each method, along with a performance comparison in the following examples:

[0104] The interpolation method uses a third-order or higher-order interpolation polynomial, which is expressed as:

[0105] P(x)=a3x 3 +a2x 2 +a1x+a0

[0106] Among them, a0, a1, a2, and a3 can be fitted by the measured value of the optical fiber loss or the quantum state amplitude in the adjacent time period. If in the embodiment, the interpolation window length is set to 2 to 4 sampling points, the cubic interpolation polynomial can maintain a prediction error of about 3% in a low vibration scenario. If the window is too large (such as ≥10 sampling points), the accuracy of small local anomalies will decrease; conversely, if the window is too small (1 sampling point), the interpolation polynomial cannot be solved. Experiments show that when the daily temperature change of the optical fiber line does not exceed 10 degrees Celsius and the quantum state amplitude phase jump amplitude is <0.1 for 2 consecutive hours, the third-order interpolation polynomial can control the average error to about 2.5% and reduce the calculation time by 20%, which is more efficient than the conventional method of direct recursion without interpolation.

[0107] In the recursive iterative method, the fractal basis function is used to perform multiple rounds of extrapolation on the loss prediction value of the previous moment and the current quantum state amplitude data (including phase coupling). It can be written as:

[0108]

[0109] Among them F k is the estimate of the loss at the next moment after the k-th round of extrapolation, ψ(t) represents the quantum state amplitude, E(t) represents the environmental parameters (such as temperature T, humidity H), is a fractal iterative mapping function. For example, in the embodiment, the iteration depth is set to 5 to 10 rounds. A learning rate below 0.001 will cause a 50% drop in convergence speed, while a learning rate above 0.1 will cause divergence. The measured optimal range is 0.01 to 0.05. Comparative results show that when fiber attenuation exhibits a sustained transition (e.g., a phase shift ≥ 0.2 for one hour), the recursive iteration method can maintain an error within 3.5%, while the error of third-order interpolation is approximately 5%, but the recursive iteration operation takes approximately 15% longer.

[0110] In terms of technical effects, the interpolation method operates more efficiently under normal or stable attenuation conditions, and there is no need to perform multiple rounds of iterations; however, it will lose precision in extreme scenarios (such as loose multi-point joints caused by construction). At this time, the recursive iteration method can capture the nonlinear fluctuations in the abnormal interval by virtue of the self-similar characteristics of fractal iteration, and improve the fault detection rate by about 10% compared with the traditional single-step extrapolation. In multiple field tests or simulation comparisons, when high temperature and stress interference are superimposed (quantum state amplitude phase jump ≥ 0.2), recursive iteration can accurately identify local faults and avoid over-simplification; once the attenuation curve returns to stability, it automatically switches back to the interpolation method to reduce unnecessary iteration overhead.

[0111] By combining interpolation with recursive iteration, this method achieves both high efficiency and high precision in fiber loss prediction, meeting the multi-scenario attenuation monitoring needs of base stations and long-distance backbone lines. Compared to existing solutions using fixed algorithms (such as third-order interpolation or single rounds of iteration), this method significantly improves overall accuracy and resource utilization. Key parameters such as the interpolation window and number of iterations are publicly adjustable, and critical value tests for the learning rate and threshold are included to ensure the method's replicability and scalability.

[0112] Preferably, during interpolation or recursive iterative extrapolation prediction, quantum state amplitude data and external environmental parameters are simultaneously input into the fractal basis function to form a multi-dimensional input condition to evaluate the optical fiber loss value at the target moment.

[0113] In this invention, during interpolation or recursive extrapolation prediction, quantum state amplitude data and external environmental parameters are input into the fractal basis function to form multidimensional input conditions, allowing for accurate assessment of fiber loss at the target time. Quantum state amplitude data typically contains a composite quantity reflecting the degree of optical field phase coupling and loss information, while external environmental parameters (such as temperature, humidity, and geological codes) can reveal the causes of nonlinear attenuation. The specific process is as follows:

[0114] First, the quantum state amplitude and environmental data within the time interval required for interpolation or recursive iteration are packaged into vector form:

[0115] X(t)=(ψ(t),E1(t),E2(t),…)

[0116] Where ψ(t) represents the quantum state amplitude (with fiber phase component), E1(t), E2(t) and so on represent environmental parameters (such as temperature T, humidity H, geological elements G). Then X(t) is sent to the fractal basis function Perform multi-dimensional mapping to obtain the attenuation estimate at the target time If the interpolation method is used, the X values of the nearest time or nearest segment are usually taken to perform cubic polynomial fitting and then extrapolate; if the recursive iterative method is used, the input is updated round by round, so:

[0117]

[0118] The fiber attenuation trend is continuously approached in multiple rounds of iterations.

[0119] In this embodiment, a window length of 5 is selected, and five sets of X(t) data are collected within each window using cubic interpolation. If a quantum state phase fluctuation δ(t) exceeding 0.1 or an external temperature change of >10°C is detected, the algorithm automatically switches to recursive iteration (depth ≥8 rounds), enabling the fractal basis function to capture stronger nonlinear fluctuations. Field observations show that for a 30-kilometer optical fiber line, moderate daily attenuation due to temperature and vibration is no longer falsely reported. Compared to a fixed cubic interpolation method, the false alarm rate is reduced by approximately 15%.

[0120] In this embodiment, if the geological code indicates a fragile rock formation (code = 1), the recursive iteration depth is increased to 10 and the learning rate η is adjusted to 0.01. In measured data, a learning rate < 0.001 results in a 50% decrease in convergence speed, while a learning rate > 0.1 results in a divergence of the curve. Within this critical range, the prediction error stabilizes at approximately 3.5%. Furthermore, if the external temperature is ≥ 40°C and the humidity is ≤ 30%, which triggers high stress attenuation, the multi-dimensional input model automatically focuses the fractal basis function on regions with large phase shifts, improving detection accuracy by approximately 5% to 7% during bursts.

[0121] In terms of technical effects, compared with existing solutions that only use single-dimensional input (such as only temperature or only quantum state amplitude), after inputting the quantum state amplitude together with external environmental parameters into the fractal basis function, short-term fluctuations and long-term trends can be captured at the same time; the most obvious gain in the experimental scenario is the ability to multi-dimensionally model when multiple disturbances (temperature surge + geological slip) occur, with an accuracy improvement of about 10%, enabling fine correction and prediction of the free curve shape. This multi-dimensional input condition can be reused in small-scale deployments (such as a single base station) and long-distance backbone lines. It is only necessary to adjust the interpolation window length or the number of iteration layers accordingly, and strictly control the learning rate and parameter threshold within the aforementioned critical range to prevent overfitting or divergence. In summary, this method can fully guarantee the flexibility and robustness of dynamic prediction of optical fiber loss, and its performance advantages in multi-fluctuation scenarios have also been verified in comparative tests.

[0122] like Figure 5 As shown, a dynamic prediction system for optical fiber loss based on free curve morphology is used to implement the dynamic prediction method for optical fiber loss based on free curve morphology. The system includes:

[0123] The wavefront phase module is used to perform wavefront tomography using at least two lenses to obtain interference patterns, and perform phase inversion and data fusion through inverse Fourier transform to obtain fused wavefront data; the wavefront phase module can use a multi-lens interferometer device combined with a high-frame rate CCD or CMOS sensor, and the interference fringe signals simultaneously collected by at least two lenses are photoelectrically converted and then handed over to an embedded processor or industrial computer to perform inverse Fourier transform; at the hardware level, a high-speed ADC (analog-to-digital converter) is usually configured for collaborative processing with an FPGA / CPU to complete phase demodulation and fuse multi-channel wavefront data in real time.

[0124] A quantum conversion module is used to align the fused wavefront data with the optical fiber loss timing data and convert them into quantum state amplitude data for use in fractal manifold construction. In hardware, the quantum conversion module is mainly embodied as an efficient data bus (such as PCIe) that inputs the fused wavefront data into a central processing unit (such as a CPU or GPU cluster) or a dedicated DSP module, and synchronously caches the timing data of an optical fiber loss monitor (such as an OTDR or power meter). After aligning the two through a high-bandwidth shared memory, information such as phase and loss is packaged into a quantum state amplitude vector, providing complex data input for subsequent iterations.

[0125] The fractal topology module is used to iteratively compute the quantum state amplitude data and external environmental parameters based on fractal basis functions to generate multi-resolution fractal manifold data. This module then extracts topological fingerprint data through topological data analysis. This module typically uses a GPU or multi-core CPU with strong parallel computing capabilities to perform fractal basis function iteration and topological analysis. The hardware can be pre-configured with a large amount of RAM to store the mesh or point cloud structure of the multi-resolution fractal manifold. After the computation is complete, the topological fingerprint data is extracted using a persistent coherence algorithm, which can be combined with double-precision floating-point operations to ensure numerical stability even at high iteration levels.

[0126] The chaos prediction module performs chaotic iterative local correction based on local anomalies detected in the topological fingerprint data and phase transitions in the quantum state amplitude data. It also generates or merges branches in the multi-resolution fractal manifold data. If a local anomaly fails to converge within a continuous time period, it determines that a new branch is needed. The fiber loss at the target time is then extrapolated and predicted through interpolation or recursive iteration. The chaos prediction module can be implemented on the same computing platform or independently using another accelerator card (such as an NVIDIA GPU or an ARM-based deep learning unit). It performs chaotic iteration, branch management, and extrapolation prediction on the input topological fingerprint and quantum state amplitude. If continuous anomalies are detected, separate branches are quickly split off using GPU acceleration for local surface correction. If the anomaly disappears, the branch results are merged back into the main manifold through data shared memory. Finally, the fiber loss prediction value generated by interpolation or recursive iteration is returned to the upper-level monitoring system.

[0127] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware.

[0128] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for dynamic prediction of optical fiber loss based on free curve morphology, characterized in that: The following steps are involved: Wavefront tomography is performed using at least two lenses to obtain an interference pattern, and phase inversion and data fusion are performed using inverse Fourier transform to obtain fused wavefront data. The phase inversion method obtains the light field amplitude and phase distribution based on interference fringe analysis and constructs a multi-channel numerical matrix. The fused wavefront data is aligned with the optical fiber loss time series data and converted into quantum state amplitude data, which is used to construct a fractal manifold. Iterative calculations are performed on quantum state amplitude data and external environment parameters using fractal basis functions, which are most suitable for scenarios with multi-scale and nonlinear attenuation characteristics. Topological fingerprint data is then extracted based on topological data analysis. Based on the local anomalies detected in the topological fingerprint data and the phase jump of the quantum state amplitude data, chaotic iterative local correction is performed to generate or merge branches in the multi-resolution fractal manifold data. When it is detected that the local anomaly has not converged within a continuous time period, it is determined that a new branch needs to be created. After correction, interpolation or recursive iteration is used to extrapolate and predict the fiber loss at the target time.

2. The method according to claim 1, characterized in that Before obtaining the interference pattern, coupling ports are set at at least two positions of the optical fiber cross section to synchronously collect light field signals, and background noise and high-frequency fringes are eliminated before performing inverse Fourier transform.

3. The method according to claim 1, characterized in that When obtaining the amplitude and phase distribution of the light field, the phase inversion method processes the light intensity component and the phase component into two-dimensional matrices respectively, and integrates at least two matrices into a multi-channel numerical matrix to characterize the local attenuation change of the optical fiber.

4. The method according to claim 1, wherein The external environmental parameters include temperature, humidity and geological factors. The fractal basis function dynamically adjusts the number of iteration layers or thresholds based on the external environmental parameters during iterative calculation to adapt to the multi-scale nonlinear attenuation characteristics.

5. The method according to claim 1, wherein Topological data analysis uses persistent homology operations to extract zero-order and first-order homology information from multi-resolution fractal manifold data to identify local anomalies corresponding to connected regions and closed loops.

6. The method according to claim 1, characterized in that The chaotic iterative local correction is based on the Logistic map or the Henon map, and the initial value of the chaos is set by the phase jump amplitude of the quantum state amplitude data.

7. The method according to claim 1, characterized in that When generating or merging branches in multi-resolution fractal manifold data, the new branch is determined by evaluating whether the local anomaly remains unconverged in consecutive time periods, and the branch is merged into the original branch after the local anomaly disappears.

8. The method according to claim 1, characterized in that The interpolation method uses at least third-order interpolation polynomials to predict the optical fiber loss at the target time, and the recursive iterative method performs multiple rounds of extrapolation operations based on fractal basis functions.

9. The method according to claim 1, characterized in that During interpolation or recursive iterative extrapolation prediction, the quantum state amplitude data and external environmental parameters are simultaneously input into the fractal basis function to form a multi-dimensional input condition to evaluate the optical fiber loss value at the target moment.

10. A system for dynamically predicting optical fiber loss based on a free curve morphology, for implementing the method for dynamically predicting optical fiber loss based on a free curve morphology according to any one of claims 1 to 9, characterized in that: The system includes: A wavefront phase module is used to perform wavefront tomography using at least two lenses to obtain an interference pattern, and perform phase inversion and data fusion through inverse Fourier transform to obtain fused wavefront data; A quantum conversion module, configured to align the fused wavefront data with the optical fiber loss timing data and convert them into quantum state amplitude data for use in fractal manifold construction; A fractal topology module is used to perform iterative operations on the quantum state amplitude data and external environment parameters based on fractal basis functions to form multi-resolution fractal manifold data, and to extract topological fingerprint data through topological data analysis; The chaos prediction module is used to perform chaotic iterative local correction based on local anomalies detected in the topological fingerprint data and phase jumps in the quantum state amplitude data, and to generate or merge branches in the multi-resolution fractal manifold data. When it is detected that the local anomaly has not converged within a continuous time period, it is determined that a new branch needs to be created. Subsequently, the fiber loss at the target time is extrapolated and predicted through interpolation or recursive iteration.

Images