High-power Raman amplifier overexcitation protection system and method

Abstract

This application provides a high-power Raman amplifier over-excitation damage monitoring and protection system and method, relating to the field of optoelectronic devices; the method includes: acquiring a first parameter sequence characterizing the pump light state and a second parameter sequence characterizing the backscattered light state; calculating a first characteristic value characterizing the spectral uniformity of the pump light based on the first parameter sequence; calculating a second characteristic value characterizing the time-domain statistical properties related to random interference of end-face reflection based on the second parameter sequence; weighting and combining the first and second characteristic values ​​to generate a real-time risk index; and selecting and executing graded protection actions according to the risk index; this application can comprehensively evaluate damage precursors from multiple physical sources, achieving early and accurate prediction and dynamic intelligent protection of damage risks, avoiding the limitations of traditional single-parameter hard threshold judgment, and significantly improving the amplifier's operational reliability and system availability.

Description

Technical Field

[0001] This application relates to the field of optoelectronic devices, and more specifically, to a high-power Raman amplifier over-excitation damage monitoring and protection system and method. Background Technology

[0002] High-power Raman amplifiers (HPAs), as key optical signal amplification devices in fiber optic communication systems, achieve gain amplification through the Raman scattering effect of pump light and signal light, and are widely used in long-distance optical transmission networks. In practical applications, high-power pump light often leads to excessive damage to the fiber endface, such as endface heat accumulation, enhanced reflection, and material degradation. This not only affects the performance stability of the amplifier but may also cause system failures or safety hazards. Existing technologies typically employ methods based on power level monitoring or simple reflection signal detection to identify damage risks. For example, backscattered light signals are captured using an optical time-domain reflectometer (OTDR), and the pump power is adjusted or shutdown protection is triggered accordingly. However, these methods mainly rely on monitoring a single parameter, such as reflection intensity or power threshold, which is insufficient to comprehensively capture the dynamic characteristics of pump light spectral inhomogeneity and random interference of endface reflections, resulting in insufficient monitoring accuracy and the inability to quantify the damage risk level in real time. Furthermore, existing protection response mechanisms are mostly static threshold triggers, exhibiting significant response lag. In high-power transient scenarios, misjudgment or protection delays are prone to occur, thus failing to effectively prevent excessive endface damage.

[0003] Therefore, there is a need for a method that can more accurately monitor the risk of end-face overexcitation damage in high-power Raman amplifiers and achieve dynamic protection, in order to overcome the shortcomings of insufficient monitoring accuracy and response lag in the existing technology. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this application provides a high-power Raman amplifier over-excitation damage monitoring and protection system and method.

[0005] Firstly, this application provides a method for protecting high-power Raman amplifiers from over-excitation damage monitoring, including:

[0006] Obtain a first real-time parameter sequence and a second real-time parameter sequence of the Raman amplifier, wherein the first real-time parameter sequence represents the state of the pump light and the second real-time parameter sequence represents the state of the backscattered light;

[0007] Based on the first real-time parameter sequence, a first characteristic value characterizing the uniformity of the pump light spectrum is calculated;

[0008] Based on the second real-time parameter sequence, a second characteristic value characterizing the time-domain statistical properties of the backscattered light is calculated, the time-domain statistical properties being related to the random interference between the pump light and the optical fiber end-face reflected light;

[0009] A real-time risk index representing the current end-face damage risk level is generated by weighting and combining the first feature value and the second feature value.

[0010] In response to the real-time risk indicator meeting the preset trigger condition, a protective action corresponding to the trigger condition is selected from a set of actions including at least two protective actions.

[0011] Optionally, the calculation of the first characteristic value characterizing the uniformity of the pump light spectrum includes:

[0012] The first real-time parameter sequence is modeled as a corresponding optical fluid velocity field;

[0013] The power spectral density of the pump light spectrum is obtained by solving the evolution of the photofluid velocity field based on the turbulence model.

[0014] A first characteristic value characterizing spectral uniformity is determined based on the power spectral density, wherein the first characteristic value is a turbulent kinetic energy dissipation index calculated based on the power spectral density.

[0015] Optionally, the calculation of the first characteristic value characterizing the uniformity of the pump light spectrum further includes:

[0016] Monitor the thermal parameters of the optical fiber end face to determine the end face boundary conditions of the photofluid velocity field;

[0017] Based on the changes in the end-face boundary conditions, the solution process of the turbulence model is modified, or the calculated turbulent kinetic energy dissipation index is corrected.

[0018] Optionally, determining the end-face boundary conditions of the photofluid velocity field includes:

[0019] An optical path coupling device with a preset splitting ratio is used to separate the local oscillating light from the forward-propagating pump light, and the local oscillating light is coherently mixed with the backscattered light to generate a mixed beam.

[0020] The mixed beam is fed into a balanced photodetector or a single-ended photodetector to generate an electrical signal characterizing the beat frequency information between the local oscillating light and the backscattered light.

[0021] The electrical signal is subjected to spectral analysis to identify and measure the power of the beat frequency component corresponding to anti-Stokes Raman scattering, or the center frequency drift of the beat frequency component corresponding to stimulated Brillouin scattering, and the power or the center frequency drift is used as the end face thermal parameter.

[0022] The end-face boundary conditions of the turbulence model are updated based on the end-face thermal parameters.

[0023] Optionally, the calculation of the second eigenvalue characterizing the temporal statistical properties of the backscattered light includes:

[0024] The backscattered light corresponding to the second real-time parameter sequence is sent into a photodetector to generate an electrical signal corresponding to the intensity of the backscattered light.

[0025] The amplitude of the electrical signal is sampled within a preset time window to obtain an amplitude sample sequence;

[0026] Calculate the second and fourth central moments of the amplitude sample sequence;

[0027] Based on the second-order central moment and the fourth-order central moment, the kurtosis value of the electrical signal is calculated, and the kurtosis value is used as the second feature value.

[0028] Optionally, the calculation of the second eigenvalue characterizing the temporal statistical properties of the backscattered light further includes:

[0029] The round-trip time delay of the backscattered light generated at a monitoring point by the fiber end face is predetermined;

[0030] The backscattered light corresponding to the second real-time parameter sequence is sent into a photodetector to generate an electrical signal;

[0031] The autocorrelation function of the electrical signal is calculated within a preset time window to obtain the autocorrelation function curve;

[0032] Extract the peak amplitude or integral energy on the autocorrelation function curve at a time delay equal to or close to the round-trip time delay, and use the peak amplitude or integral energy as the second characteristic value.

[0033] Optionally, extracting the peak amplitude or integral energy on the autocorrelation function curve at a time delay equal to or near the round-trip time delay, and using the peak amplitude or integral energy as the second feature value includes:

[0034] Obtain the main correlation peak value on the autocorrelation function curve at the point where the time delay is zero, where the main correlation peak value represents the total backscattered light power;

[0035] Obtain the target correlation peak on the autocorrelation function curve at a time delay equal to or close to the round-trip time delay, wherein the target correlation peak represents the power of the coherent interference component caused by end-face reflection;

[0036] Calculate the ratio of the target correlation peak value to the main correlation peak value;

[0037] And the ratio is used as the second characteristic value.

[0038] Optionally, the step of generating a real-time risk indicator characterizing the current end-face damage risk level by weighting and combining the first feature value and the second feature value includes:

[0039] The current average pump power of the Raman amplifier is obtained in real time;

[0040] Based on the current average pump power, a first weight value and a second weight value corresponding to the current average pump power are determined from a preset weight mapping relationship, wherein the weight mapping relationship is configured such that the second weight value increases non-linearly with the increase of the current average pump power.

[0041] Multiplying the first feature value by the first weight value yields the first weighted feature value, and multiplying the second feature value by the second weight value yields the second weighted feature value;

[0042] The first weighted feature value and the second weighted feature value are combined to generate the real-time risk indicator.

[0043] Optionally, in response to the real-time risk indicator satisfying a preset trigger condition, selecting a protective action corresponding to the trigger condition from a set of actions including at least two protective actions includes:

[0044] The triggering conditions of the real-time risk indicator are decomposed into a first sub-condition determined based on the comparison relationship between the first feature value and the first threshold, and a second sub-condition determined based on the comparison relationship between the second feature value and the second threshold.

[0045] The set of actions includes a first type of protective action and a second type of protective action;

[0046] When the first sub-condition is met but the second sub-condition is not met, the first type of protection action is selected. The first type of protection action is designed to adjust the electronic parameters related to the pump light spectrum.

[0047] When the second sub-condition is met, the second type of protection action is selected, which includes reducing or shutting down the power of the pump source.

[0048] Secondly, this application provides a high-power Raman amplifier over-excitation damage monitoring and protection system, including:

[0049] The acquisition module is used to acquire a first real-time parameter sequence and a second real-time parameter sequence of the Raman amplifier, wherein the first real-time parameter sequence represents the state of the pump light and the second real-time parameter sequence represents the state of the backscattered light.

[0050] The first processing module calculates a first characteristic value characterizing the spectral uniformity of the pump light based on the first real-time parameter sequence; and calculates a second characteristic value characterizing the temporal statistical properties of the backscattered light based on the second real-time parameter sequence, wherein the temporal statistical properties are related to the random interference between the pump light and the light reflected from the fiber end face.

[0051] The second processing module performs a weighted combination of the first feature value and the second feature value to generate a real-time risk indicator that characterizes the current end face damage risk level.

[0052] The protection module is used to select a protection action corresponding to the trigger condition from a set of actions including at least two protection actions in response to the real-time risk indicator meeting the preset trigger condition.

[0053] Compared with existing technologies, this application, by simultaneously acquiring real-time parameter sequences of pump light and backscattered light and calculating characteristic values ​​such as spectral uniformity and temporal statistical properties, can more comprehensively capture the dynamic signals of end-face damage, achieving precise quantification of damage risk. This significantly improves monitoring accuracy and real-time performance compared to the single-parameter monitoring of existing technologies. Furthermore, by generating risk indicators through weighted combination and selecting protective actions based on triggering conditions, this invention achieves dynamic optimization of the response mechanism, avoiding the lag problem of existing technologies and ensuring timely protection against end-face damage in high-power scenarios, thereby improving the overall reliability of the amplifier and the stability of the system. Attached Figure Description

[0054] Figure 1 A flowchart illustrating the high-power Raman amplifier overexcitation damage monitoring and protection method provided in this application embodiment;

[0055] Figure 2 A flowchart illustrating a method for calculating a first eigenvalue characterizing the uniformity of the pump light spectrum, provided in an embodiment of this application;

[0056] Figure 3 A flowchart illustrating another method for calculating a first eigenvalue characterizing the uniformity of the pump light spectrum, provided in an embodiment of this application;

[0057] Figure 4 A schematic diagram of a high-power Raman amplifier overexcitation damage monitoring and protection system provided in an embodiment of this application.

[0058] Explanation of reference numerals in the attached diagram: 10, acquisition module; 20, first processing module; 30, second processing module; 40, protection module. Detailed Implementation

[0059] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0060] See Figure 1 The diagram shows a flowchart of a high-power Raman amplifier overexcitation damage monitoring and protection method provided in this application embodiment. The method includes steps S101 to S105, wherein:

[0061] S101: Obtain the first real-time parameter sequence and the second real-time parameter sequence of the Raman amplifier, wherein the first real-time parameter sequence represents the state of the pump light and the second real-time parameter sequence represents the state of the backscattered light.

[0062] S102: Based on the first real-time parameter sequence, calculate a first characteristic value characterizing the uniformity of the pump light spectrum;

[0063] S103: Based on the second real-time parameter sequence, calculate a second characteristic value characterizing the time-domain statistical properties of the backscattered light, wherein the time-domain statistical properties are related to the random interference between the pump light and the optical fiber end-face reflected light;

[0064] S104: Based on the first feature value and the second feature value, a weighted combination is performed to generate a real-time risk index characterizing the current end face damage risk level;

[0065] S105: In response to the real-time risk indicator meeting the preset trigger condition, select the protective action corresponding to the trigger condition from the action set including at least two protective actions.

[0066] Regarding the above S101:

[0067] In practical implementation, it is necessary to acquire the sequence of key parameters characterizing the internal operating state of the amplifier in real time and synchronously. Optionally, this requires appropriate modification to the optical path of the Raman amplifier to extract the monitoring signal non-invasively. Specifically, optical splitters can be set on the main optical path before the pump light of the Raman amplifier enters the gain fiber, and at the optical port used to receive backscattered light.

[0068] To obtain the first real-time parameter sequence characterizing the pump light state, this embodiment may include the following specific steps:

[0069] First, a first fiber coupler with a preset splitting ratio is fused or connected to the main optical path between the pump source output and the gain fiber input. As a specific example, this first fiber coupler can employ a 1:99 splitting ratio, meaning that one percent of the pump light power is coupled to the monitoring optical path, while ninety-nine percent of the power continues to be transmitted to the gain fiber for pump amplification, thus ensuring that the monitoring process does not significantly affect the normal operation of the amplifier.

[0070] Next, the pump optical signal output from the monitoring port of the first fiber coupler is fed into a first broadband photodetector. This detector is responsible for linearly and with high fidelity converting the received pump optical signal, whose optical power changes rapidly over time, into a continuously varying analog electrical signal, such as a voltage signal or a current signal. The instantaneous amplitude of this analog electrical signal is proportional to the instantaneous power of the pump light.

[0071] Then, the analog electrical signal output from the first broadband photodetector is input to the sampling input of a high-speed analog-to-digital converter. The high-speed analog-to-digital converter periodically samples and quantizes the analog electrical signal at a preset high sampling frequency.

[0072] Finally, the high-speed analog-to-digital converter outputs a series of discrete digital values. This sequence of digital values, arranged in chronological order, constitutes the first real-time parameter sequence as referred to in this application. Each value in this sequence precisely corresponds to the instantaneous pump light power at a sampling moment.

[0073] In other embodiments, a sequence of parameters characterizing the state of the pump light can also be obtained by monitoring the drive current of the pump laser.

[0074] To obtain the second real-time parameter sequence characterizing the backscattered light state, the process is similar to that of obtaining the first real-time parameter sequence, and may include the following specific steps:

[0075] First, the optical path of a high-power Raman amplifier typically includes an optical circulator, one function of which is to guide the backscattered light returning from the gain fiber to a specific output port. A second fiber coupler is connected to the optical path of this output port to separate a portion of the backscattered light for monitoring.

[0076] Secondly, the backscattered light signal output from the monitoring port of the second fiber coupler is fed into a second broadband photodetector to convert the real-time change in its optical power into a corresponding continuous analog electrical signal.

[0077] Then, the analog electrical signal output by the second broadband photodetector is input to a high-speed analog-to-digital converter for sampling and quantization.

[0078] Finally, the high-speed analog-to-digital converter outputs another discrete sequence of digital values, which constitutes the second real-time parameter sequence referred to in this application. Each value in this sequence corresponds to the instantaneous power of the total backscattered light at the sampling time.

[0079] Through the above steps, this method obtains two high-fidelity digital signal streams that reflect the dynamic changes of the optical field inside the amplifier. These two parameter sequences will serve as the basic inputs for subsequent risk feature extraction and assessment. For example, the first real-time parameter sequence includes information reflecting the operational stability of the pump source, which can be used for subsequent analysis of the pump source's spectral quality; while the second real-time parameter sequence includes rich information on various scattering and reflection events in the optical fiber, which can be used for subsequent analysis of abnormal signal fluctuations caused by factors such as end-face reflection.

[0080] Regarding step S102:

[0081] Step S102 calculates a first eigenvalue that can quantify the uniformity of the pump light spectrum based on the first real-time parameter sequence. The uniformity of the pump light spectrum is directly related to its ability to suppress nonlinear effects such as stimulated Brillouin scattering, and is an important dimension for assessing the potential damage risk of the amplifier, but it is usually ignored by existing technologies.

[0082] Optionally, the following processing flow may be included:

[0083] First, the first real-time parameter sequence representing the pump light state obtained in step S101 is modeled. This parameter sequence is essentially a record of high-frequency fluctuations in the light field intensity or its related quantities over time. In this embodiment, this high-dimensional time-series data can be constructed into a mathematical model through a preset mapping rule. This model can be conceptually compared to the velocity field of an optical fluid.

[0084] Specifically, the instantaneous amplitude and rate of change of an optical signal can be analogized to vector parameters such as velocity and acceleration of a fluid at a specific point in time and space, thus transforming a time-domain optical signal problem into a dynamic problem that can be analyzed within the framework of fluid mechanics.

[0085] Secondly, the constructed photofluid velocity field model is solved through evolution. In this field, a flat, uniformly broadened chaotic spectrum, whose energy is uniformly distributed in the frequency domain, can be analogized to a fully developed turbulent state in which energy is efficiently transferred in vortices at all scales.

[0086] Conversely, a non-uniform spectrum with numerous power spikes caused by hardware defects, such as clock jitter, can be analogized to an unstable "non-ideal turbulence" state where energy is abnormally concentrated at a specific scale. Therefore, this embodiment can numerically calculate or simulate the evolution of the photofluid velocity field based on a preset turbulence model. As a concrete example, a computational framework similar to classic turbulence models in fluid mechanics, such as large eddy simulation, can be used to solve for the energy distribution of the photofluid in the frequency domain, which corresponds to the power spectral density of the pump light spectrum.

[0087] Furthermore, based on the obtained power spectral density, the final first eigenvalue is calculated. To quantify the uniformity of the spectrum, this embodiment introduces a specific evaluation metric.

[0088] For example, a turbulent kinetic energy dissipation index can be used. In fluid dynamics analogy, this index characterizes the efficiency of energy transfer from large-scale eddies to small-scale eddies and its eventual dissipation. A uniform spectrum corresponds to efficient and balanced energy dissipation, while a non-uniform spectrum corresponds to inefficient and concentrated energy dissipation. By performing specific integration or statistical operations on the solved power spectral density, a scalar value quantifying this dissipation efficiency can be obtained.

[0089] Ultimately, this calculated scalar value is used as the first characteristic value characterizing the current pump light spectral uniformity. A higher characteristic value may indicate better and more uniform spectral quality; conversely, a lower value indicates deteriorated spectral quality and potential risks. The calculated first characteristic value will be output to the subsequent risk assessment to characterize the potential risks from the pump source side.

[0090] Those skilled in the art should understand that the above-described calculation method based on fluid dynamics analogy is only an optional embodiment of this application. In other embodiments, other signal processing methods well known to those skilled in the art can also be used to calculate the first feature value. For example, the power spectrum can be obtained by directly performing a fast Fourier transform on the first real-time parameter sequence, and then the first feature value characterizing the spectral uniformity can be obtained by calculating the peak-to-average power ratio, flatness, or statistical variance of the power spectrum.

[0091] Regarding the above S103:

[0092] Step S103 calculates a second eigenvalue that characterizes the temporal statistical properties of backscattered light based on the second real-time parameter sequence obtained in step S101. Traditional total power monitoring methods are not sensitive to instantaneous power spikes caused by random interference between chaotic pump light and fiber end-face reflected light, which are direct precursors to damage. Therefore, this step aims to extract features that accurately reflect this risk from the complex backscattered light signal through more advanced signal processing methods.

[0093] As an alternative implementation, risk can be identified by analyzing the probability distribution of signal amplitude. The basic idea is that in a healthy system containing only background noise, the signal amplitude fluctuations approximate a normal distribution; however, when instantaneous power spikes caused by random interference begin to appear, the tail of its probability density function becomes significantly "fat," meaning the probability of extremely high amplitude values ​​increases. This embodiment quantifies the risk by calculating a statistical indicator that sensitively reflects this fat-tailed effect.

[0094] In specific implementation, the second real-time parameter sequence obtained in step S101 is preprocessed. Optionally, the DC component or mean of the digital sequence can be removed first to obtain an amplitude sample sequence that only includes alternating fluctuation information, so that subsequent statistical calculations can focus more on the dynamic characteristics of the signal.

[0095] Next, the second and fourth central moments are calculated on the amplitude sample sequence. In this field, the second central moment is the variance of the signal, characterizing the average fluctuation power of the signal. The fourth central moment characterizes the kurtosis and tail characteristics of the signal probability density function distribution. The calculation process is as follows: First, the average value of the amplitude sample sequence is calculated; then, the deviation between each sample point and the average value is calculated, and the average of the square and fourth powers of these deviations is calculated respectively, thereby obtaining the second and fourth central moments of the sequence.

[0096] Finally, based on the calculated second and fourth central moments, a kurtosis value characterizing the probability distribution of the sequence is calculated using a pre-defined statistical formula. This kurtosis value can sensitively reflect the frequency of extreme outliers in the signal, i.e., instantaneous power spikes. An abnormal increase in its value directly indicates an increase in random interference and a higher risk of damage. Therefore, the calculated kurtosis value, or its associated normalized value, can be used as the second characteristic value.

[0097] As an optional implementation, risks can be identified by analyzing the inherent temporal correlation of signals. The basic idea is that random interference signals caused by reflections from the fiber optic endface necessarily include an echo characteristic corresponding to the physical location of the endface; that is, the signal contains a copy of itself after a fixed time delay. This embodiment quantifies the risk by detecting the intensity of this echo characteristic. Specifically, the processing flow may include the following:

[0098] First, a key physical parameter is predetermined: the round-trip time delay of the optical signal propagating from the monitoring point in the system to the end face of the gain fiber and reflecting back to the monitoring point. This parameter is uniquely determined by the fiber length and the core refractive index, and is an inherent, stable physical constant.

[0099] Next, the second real-time parameter sequence obtained in step S101 or its photoelectric converted electrical signal is subjected to autocorrelation function calculation within a preset time window. This calculation process evaluates the similarity between the signal sequence and itself after different time shifts, and generates an autocorrelation function curve, where the horizontal axis represents the time delay and the vertical axis represents the correlation amplitude.

[0100] Then, peak features at specific delay locations are extracted from the autocorrelation function curve. Specifically, a target correlation peak is found at a time delay equal to or near the preset round-trip time delay. The presence and intensity of this peak directly demonstrate and quantify the coherent reflection components from the fiber endface.

[0101] As an optional implementation, in order to improve the reliability of the feature value and eliminate the interference of pump power fluctuations, the extraction process further includes: firstly, obtaining the main correlation peak at the time delay equal to zero, which represents the total incoherent backscattering power reference; then, calculating the ratio of the target correlation peak at the round-trip time delay position to the main correlation peak, which is the coherent / incoherent scattering ratio.

[0102] Finally, the peak amplitude, integral energy, or ratio of the extracted target correlation peak to the main correlation peak is used as the second characteristic value characterizing the intensity of the current random interference risk.

[0103] Those skilled in the art will understand that the second characteristic value calculated through any of the above embodiments can accurately quantify the damage risk caused by random interference from end-face reflections from different statistical dimensions. This second characteristic value will then be fed into a subsequent risk assessment and fused with the first characteristic value to form a comprehensive judgment of the amplifier's health status.

[0104] Regarding S104 and S105 above:

[0105] Step S104, based on the first and second feature values ​​calculated in the previous steps, which respectively characterize two different physical risk sources, performs effective information fusion to generate a real-time risk indicator that can comprehensively and accurately assess the current overall damage risk. This step aims to overcome the limitations of existing technologies that view a single parameter in isolation, and to achieve a comprehensive judgment of multi-dimensional risk information.

[0106] In a specific embodiment, the weighted combination can be implemented using a preset linear fusion algorithm. Specifically, the processing module can pre-store fixed weight coefficients, including a first weight coefficient and a second weight coefficient. During runtime, the processing module multiplies the first feature value calculated in step S102 by the first weight coefficient, and multiplies the second feature value calculated in step S103 by the second weight coefficient. Then, the two products are arithmetically summed or other preset combination operations are performed, and the final result is the real-time risk index. The weight coefficients can be preset based on a large amount of experimental data statistics or theoretical model simulation to reflect the difference in the contribution of the two different risk sources, pump source spectral degradation and end-face reflection interference, to the final overexcitation damage.

[0107] Optionally, to more accurately simulate the nonlinear characteristics of damage risk dynamically changing with amplifier operating conditions, a dynamic adaptive weighting method can be employed. This method recognizes that different risk sources have varying degrees of danger at different power levels. Specifically, it may include the following process:

[0108] First, in addition to acquiring the first and second real-time parameter sequences, the monitoring system also acquires the current average pump power of the Raman amplifier in real time, which serves as a key parameter characterizing the current macroscopic operating state of the amplifier.

[0109] Secondly, a pre-defined weight mapping relationship is established, which can be stored in the form of a lookup table or a function expression. This mapping relationship establishes a one-to-one correspondence between the current average pump power and the first and second weight values. A core design principle of this mapping relationship is that as the pump power increases, the second weight value associated with the end-face reflection interference risk (second eigenvalue) increases non-linearly and rapidly, reflecting the physical reality that at high power, any tiny enhancement of end-face reflection can lead to catastrophic consequences.

[0110] Finally, the processing module dynamically queries or calculates the applicable first and second weight values ​​from the weight mapping relationship based on the real-time acquired current average pump power, and uses this pair of dynamic weight values ​​to perform a weighted combination operation on the first and second feature values, thereby generating a real-time risk indicator that more accurately reflects the true risk level under the current operating conditions.

[0111] Regarding step S105:

[0112] The core task of step S105 is to make intelligent and timely decisions based on the real-time risk indicators generated in step S104, and to select the most appropriate protective action. This application aims to achieve a tiered and dynamic protection strategy, rather than the one-size-fits-all emergency shutdown found in traditional technologies.

[0113] For example, multiple risk levels can be preset, each defined by one or more risk thresholds. The protection module compares the real-time risk indicators generated in S104 with these preset risk thresholds in real time. Simultaneously, at least two protective actions of different intensities are predefined in the system's action set.

[0114] For example, a lower warning threshold and a higher danger threshold can be set. When the real-time risk indicator exceeds the warning threshold but is below the danger threshold, the system selects to execute the first-level protection action, such as sending an alarm message to the network management system via a digital interface and controlling the pump source driver to moderately reduce its output power, for example, by 10%. When the real-time risk indicator exceeds the danger threshold, the system selects to execute the highest-level protection action, such as urgently shutting down the pump source within microseconds and triggering a local audible or visual alarm. After selecting the action, the protection module immediately generates control commands and adjusts the relevant operating parameters of the Raman amplifier through the drive circuit to execute the protection action.

[0115] Optionally, in order to achieve differentiated and precise protection based on fault tracing, the selection of protection actions not only depends on the total score of the final real-time risk indicator, but also traces back to the independent first and second feature values ​​that constitute the indicator.

[0116] Specifically, the preset triggering conditions can be decomposed into sub-conditions corresponding to the two eigenvalues. For example, a two-dimensional decision matrix can be preset. When only the first eigenvalue exceeds its corresponding sub-threshold, the system determines that the risk mainly originates from the pump source itself. In this case, it selects to execute the first type of protection action, which aims to repair or avoid electronic problems. For example, it triggers the reset procedure of the chaotic signal generator in the pump source drive circuit, or sends a specific alarm requiring pump source maintenance. When the second eigenvalue exceeds its corresponding sub-threshold, the system determines that there is an imminent risk of physical damage. In this case, it selects to execute the second type of protection action, which aims to directly reduce optical power, such as urgently reducing pump power or immediately shutting down. When both eigenvalues ​​exceed the limit simultaneously, the highest level of combined shutdown and alarm action is executed.

[0117] Optional, see Figure 2 The flowchart of a method for calculating a first eigenvalue characterizing the uniformity of the pump light spectrum provided in this application embodiment includes steps S201 to S203, wherein:

[0118] S201: Model the first real-time parameter sequence as a corresponding optical fluid velocity field;

[0119] S202: The power spectral density of the pump light spectrum is obtained by solving the evolution of the photofluid velocity field according to the turbulence model;

[0120] S203: Determine a first characteristic value characterizing the spectral uniformity based on the power spectral density, wherein the first characteristic value is a turbulent kinetic energy dissipation index calculated based on the power spectral density.

[0121] Research has found that in real-time protection scenarios involving high-power Raman amplifiers, traditional methods that directly measure the spectrum using instruments such as optical spectrum analyzers suffer from fundamental drawbacks, including slow response times, high costs, and difficulty in integration, failing to meet the real-time requirements of transient damage protection. Therefore, a novel method is needed that can indirectly, rapidly, and at low cost assess spectral uniformity.

[0122] This application introduces an interdisciplinary modeling approach, transforming the complex problem of time-domain optical signal analysis into a dynamic problem that can be efficiently solved within a fluid dynamics framework.

[0123] In this way, by performing in-depth numerical calculations on the first real-time parameter sequence, physical features strongly correlated with spectral uniformity can be extracted with high precision without relying on any optical spectrum measurement hardware, thereby achieving rapid, online diagnosis of the pump source spectral quality.

[0124] In specific implementation, the first real-time parameter sequence acquired in S101 is modeled as a corresponding photofluid velocity field. The first real-time parameter sequence is a series of discrete digital values ​​arranged in time sequence, representing the instantaneous power of the pump light. For dynamic analysis, it needs to be mapped into a multi-dimensional mathematical model.

[0125] For example, the instantaneous power value at the sampling moment can be taken as one component of the velocity field, and the first-order time derivative of the instantaneous power value, i.e., the rate of change, can be taken as a second component. Higher-order time derivatives can also be introduced as components with more dimensions. In this way, a one-dimensional time series is mapped to a multi-dimensional vector field evolving in a virtual computing space. This vector field is defined as the photofluid velocity field referred to in this application, which contains the amplitude and dynamic change information of the original optical signal.

[0126] Secondly, the photofluid velocity field is evolved and solved according to the turbulence model to obtain the power spectral density of the pump light spectrum. In the physical analogy of this application, a uniformly broadened ideal chaotic spectrum, whose energy is uniformly dispersed in the frequency domain, can be analogized to a fully developed, stable turbulent state in which energy is efficiently transferred in vortices at all scales. Conversely, a non-uniform spectrum with a large number of power spikes can be analogized to an unstable turbulent state in which energy is abnormally concentrated at a specific scale.

[0127] Therefore, this embodiment uses a preset turbulence model to numerically evolve and solve the photofluid velocity field constructed in the previous step.

[0128] For example, mature computational frameworks in computational fluid dynamics, such as the core ideas in large eddy simulation (LES) or Reynolds-mean-stress (RANS) models, can be used to simulate the transfer and distribution of energy in the photofluid at different scales, i.e., between frequencies. The final steady-state solution of this simulation evolution process, or its solution at a specific time step, will give a function characterizing the scale distribution of turbulent energy, which is physically equivalent to the power spectral density of the original pump light signal.

[0129] Then, a first eigenvalue characterizing the spectral uniformity is determined based on the obtained power spectral density. In order to quantify the uniformity of the entire power spectral density with a single scalar, this embodiment introduces a specific evaluation index, namely the turbulent kinetic energy dissipation index.

[0130] In fluid dynamics analogy, this index characterizes the efficiency of energy transfer and dissipation from large-scale vortices to small-scale vortices. A uniform spectrum corresponds to efficient and balanced energy dissipation. Specifically, a weighted integral or statistical operation can be performed on the obtained power spectral density curve. For example, a higher weight can be assigned to the power spectral density in the high-frequency portion, because the effective dissipation of energy in these high-frequency vortices better represents the uniform broadening of the spectrum. The result of this weighted integral or statistical operation is defined as the turbulent kinetic energy dissipation index and serves as the final first characteristic value.

[0131] Thus, this embodiment transforms the complex problem of spectral quality assessment into a numerical computation problem that can be efficiently executed in a digital signal processor or field-programmable gate array. The final calculated first feature value will be output to the risk assessment module in step S104, providing a key input dimension characterizing the health status of the pump source for subsequent comprehensive decision-making.

[0132] For example, to ensure compatibility with system hardware, the first real-time parameter sequence can be acquired by a high-speed sampling device, with a sampling rate preferably not less than 10 GS / s, and a sliding time window length of 0.4–4 µs (corresponding to 2 10 ~2 14 (Sampling points), to cover events of Brillouin linewidth magnitude while taking into account implementation complexity.

[0133] For example, the first real-time parameter sequence undergoes DC component removal, amplitude normalization, and slow-varying trend elimination. If necessary, fixed-bandwidth digital filtering is used to suppress electrical noise and power frequency interference. Subsequently, at least three types of quantities are extracted at each sampling point: instantaneous amplitude, rate of change between adjacent sampling points (discrete first-order difference), and curvature (discrete second-order difference), and these three are arranged into a vector in a fixed order. Within a sliding time window, the above vector sequence is mapped onto a one-dimensional uniform grid to form the "initial field" of the window; the number of grid points can be 1024 to 4096. To ensure the consistency of subsequent spectral quantity calibration, a reference frequency range covering the Brillouin linewidth is set as the calibration basis for scale mapping.

[0134] For example, a one-dimensional nonlinear convection-diffusion turbulence model is selected, and a subgrid viscosity term related to the local gradient is introduced to suppress numerical oscillations and simulate small-scale dissipation; the subgrid constant can be taken as 0.10 to 0.25. Natural or mixed boundary conditions are adopted at both ends of the virtual space. The boundary parameters can be determined by the end-face thermal parameters, such as temperature, normal heat flux, heat transfer coefficient, or equivalent heat capacity of the end face, to reflect the influence of the end-face state on the spectral evolution. These end-face parameters can be obtained and dynamically updated according to other embodiments of this application. Spatial discretization can adopt a second-order central difference or equivalent precision format; time progression adopts a multi-step method or semi-implicit method with decreasing total variation characteristics. The time step is selected according to conventional numerical stability constraints to ensure that the convection and diffusion terms do not cross multiple grid points within a single step. Iterations are performed for 50 to 200 steps until the field quantity change is lower than a set threshold or reaches a fixed step limit. After reaching a quasi-steady state or a specified iteration step, a discrete Fourier transform is performed on the field distribution in the virtual space to obtain the energy distribution of each "spatial mode". Combined with the aforementioned calibration relationship, the mode index is mapped to the actual frequency to form an estimate of the pump power spectral density. If necessary, a window function or a fixed-resolution bandwidth moving average is applied to suppress sidelobes and discretization errors.

[0135] For example, a "turbulent kinetic energy dissipation index" is calculated as the first eigenvalue based on the obtained power spectral density. This index is used to measure spectral uniformity as a single scalar. During calculation, higher weights are assigned to frequency bands above the reference frequency (corresponding to orders of magnitude of the Brillouin linewidth), making the index more sensitive to narrow spectral peaks and fine-scale ripples. The weighting index can be taken from 1 to 3 to adapt to different operating conditions and sensitivities. To enhance robustness, the spectral curve can be median filtered or segmented to remove extreme values, and a moving average can be performed over a fixed resolution bandwidth. The resulting index output can be linearly normalized to the 0-1 interval, facilitating fusion with the second eigenvalue and threshold management.

[0136] Optionally, the sampling rate is 10 GS / s, the window length is 4096 points, the step size is 1–10 µs, the number of grid points is 1024, and the subgrid constant is 0.18. When a change in end-face temperature rise or equivalent absorption increment is detected, the model boundary parameters are updated synchronously and the solution is recalculated in the next window to reflect the impact of end-face state changes on spectral uniformity. The above process can be implemented in a pipelined manner on a digital signal processor or field-programmable gate array, meeting the online monitoring requirements without significantly increasing hardware complexity.

[0137] It is understood that the photofluid velocity field and turbulence model referred to in this application are computational abstractions for rapidly estimating spectral uniformity, intended to realize the redistribution and measurement of energy across scales on digital processors or programmable logic devices, and are not limited to real fluid processes.

[0138] Optional, see Figure 3 The flowchart illustrates another method for calculating a first eigenvalue characterizing the uniformity of the pump light spectrum, provided in an embodiment of this application, including steps S301 to S302, wherein:

[0139] S301: Monitor the thermal parameters of the optical fiber end face to determine the end face boundary conditions of the photofluid velocity field;

[0140] S302: Based on the changes in the end-face boundary conditions, the solution process of the turbulence model is modified, or the calculated turbulent kinetic energy dissipation index is corrected.

[0141] Research has revealed that the optofluidic model constructed in the above embodiments is, in its basic form, an open-loop system. It does not consider the dynamic physical effects of high-power laser light on the fiber medium, i.e., the model's boundaries. At high power, thermal effects inevitably occur at the fiber endface, causing changes in physical parameters such as local refractive index. These changes, in turn, affect the propagation and spectral morphology of the pump light. A model that does not consider this real physical feedback will gradually decrease in computational accuracy as power increases and time progresses.

[0142] This application establishes a closed-loop, adaptive computational model. By monitoring the thermal parameters of the fiber end face in real time and feeding them back into the model as dynamic boundary conditions of the optofluid model, the model's evolution and solution can perceive the real changes in the physical world in real time and thus perform self-correction.

[0143] This greatly improves the accuracy and reliability of the first characteristic value calculation under various power levels and operating environments, making the early warning capability of the entire monitoring and protection system more robust and accurate.

[0144] In practice, the thermal parameters of the optical fiber end face are monitored to determine the end face boundary conditions of the photofluid velocity field. This step aims to obtain a physical quantity that can reflect the thermal state of the end face in real time.

[0145] For example, temperature can be directly measured by placing a highly sensitive temperature sensor, such as a miniature thermistor or thermocouple, on the connector housing near the fiber optic end face. Alternatively, to avoid physical contact and thermal conduction delays, a non-contact optical measurement method can be used to obtain optical parameters directly related to the end face thermal state.

[0146] The monitored thermal parameters, such as a temperature value, will be used to determine the boundary conditions of the optofluid model at the physical interface of the fiber end face. In fluid dynamics analogy, this is equivalent to determining key parameters such as "wall temperature," "wall roughness," or "fluid viscosity" at the pipe outlet. These parameters directly affect the fluid behavior at the boundary; for example, an increased "wall temperature" will introduce additional thermal disturbances or alter flow characteristics in the model.

[0147] Secondly, based on the changes in the end-face boundary conditions, the solution process of the turbulence model is modified, or the calculated turbulent kinetic energy dissipation index is corrected. This step has two possible implementation paths:

[0148] One implementation involves making corrections during the solution process. Specifically, the real-time updated end-face boundary conditions are treated as a dynamic input parameter and directly fed back into the numerical calculation or simulation evolution of the turbulence model. This means that at each computation time step, the model solution takes into account the latest end-face thermal state, resulting in a power spectral density that already includes the influence of thermal effects, making the calculation results more accurate and closer to physical reality.

[0149] Another implementation involves post-solution correction. Specifically, an initial turbulent kinetic energy dissipation index can be calculated initially, disregarding thermal effects. Then, based on the monitored thermal parameters, the system searches for or calculates a corresponding correction coefficient or correction amount from a pre-defined correction function or correction lookup table. Finally, this correction coefficient or correction amount is used to compensate for or correct the initial turbulent kinetic energy dissipation index, yielding the final first characteristic value. This correction function or lookup table can be pre-established based on experimental data calibration or theoretical modeling.

[0150] By introducing this closed-loop adaptive correction mechanism based on real physical feedback, the first eigenvalue calculated in this embodiment can more accurately reflect the true spectral state of the amplifier under high power thermal load, greatly improving the early warning reliability of the entire monitoring and protection system. The corrected first eigenvalue will be sent to the risk assessment module in step S104 along with other eigenvalues, providing a more accurate basis for the system to make a final decision.

[0151] Optionally, determining the end-face boundary conditions of the photofluid velocity field includes:

[0152] An optical path coupling device with a preset splitting ratio is used to separate the local oscillating light from the forward-propagating pump light, and the local oscillating light is coherently mixed with the backscattered light to generate a mixed beam.

[0153] The mixed beam is fed into a balanced photodetector or a single-ended photodetector to generate an electrical signal characterizing the beat frequency information between the local oscillating light and the backscattered light.

[0154] The electrical signal is subjected to spectral analysis to identify and measure the power of the beat frequency component corresponding to anti-Stokes Raman scattering or the center frequency drift of the beat frequency component corresponding to stimulated Brillouin scattering, and the power or the center frequency drift is used as the end face thermal parameter.

[0155] The end-face boundary conditions of the turbulence model are updated based on the end-face thermal parameters.

[0156] Research has found that traditional methods for monitoring thermal parameters, such as placing physical sensors like thermistors near the fiber endface, suffer from fundamental drawbacks including significant response delays, indirect measurement, and low spatial resolution. Heat transfer from the micrometer-scale fiber core to the millimeter-scale sensor takes time, causing the monitored signal to lag behind the actual instantaneous temperature rise. Furthermore, the sensors measure the average temperature of macroscopic components such as connectors, failing to accurately reflect localized hotspots within the fiber core itself. These limitations render traditional temperature measurement methods inadequate for meeting the stringent requirements of effective early warning and protection against transient over-excitation damage.

[0157] This embodiment employs a physical concept of self-differential beat coherent detection to achieve non-contact, in-situ, and ultrafast optical measurement of end-face thermal parameters. Its basic principle is to utilize a small portion of the forward-propagating pump light itself as a high-quality "reference light," i.e., local oscillating light, to coherently interfere with the backscattered light carrying end-face thermal state information.

[0158] In this way, extremely weak frequency and power changes in the optical signal related to temperature can be converted with high sensitivity into the frequency domain of electrical signals that are easily processed by electronics. The beneficial effect is that it realizes an optical thermometer with a response speed on the order of nanoseconds that directly reflects the physical state of the fiber core, completely overcoming the delay and accuracy bottlenecks of traditional physical sensors, and providing a data foundation for the accuracy and real-time performance of the entire protection model.

[0159] In a specific implementation, an optical path coupling device with a preset splitting ratio is used for coherent mixing. For example, this optical path coupling device can be a 2x2 fiber coupler. One input port of the coupler receives a beam from the pump source, for example, split from the first fiber coupler as described in claim 1, as a local oscillator. The other input port receives the complete backscattered light from the gain fiber via an optical circulator. The local oscillator and the backscattered light interfere within the coupler, i.e., coherently mix, and two mixed beams with different phases are output from the two output ports of the coupler.

[0160] Optionally, to improve mixing efficiency, a polarization controller can be added to the optical path of the local oscillating light or the backscattered light to match their polarization states.

[0161] Furthermore, the mixed beams are photoelectrically converted. For example, a balanced photodetector is used. The two mixed beams output from the coupler are input to the two photosensitive input terminals of the balanced photodetector. The balanced detector performs photoelectric conversion and differential amplification on the two optical signals. This method effectively suppresses the intensity noise of the local oscillation light itself, thereby greatly improving the signal-to-noise ratio of the beat frequency signal. Optionally, a conventional single-ended high-speed photodetector can also be used to receive one of the mixed beams for photoelectric conversion. Regardless of the method used, the detector ultimately outputs an analog electrical signal whose spectrum includes the beat frequency information between the various frequency components of the local oscillation light and the backscattered light.

[0162] Furthermore, spectral analysis is performed on the electrical signal to extract thermal parameters. For example, a digital signal processor performs high-speed sampling and spectral analysis on the electrical signal, for instance, through a Fast Fourier Transform unit. This step provides two alternative, preferred paths for extracting thermal parameters:

[0163] Path 1 based on anti-Stokes power: Because anti-Stokes Raman scattered light has a specific and significant frequency upshift relative to the pump light (local oscillator light), its beat frequency signal will appear at a specific high-frequency position in the electrical signal spectrum. The system sets an analysis window at this specific frequency position and measures the signal power or integrated energy within the window. This measured value is positively correlated with the fiber temperature and can be directly used as the end-face thermal parameter.

[0164] Based on Stimulated Brillouin Scattering (SBS) frequency drift path two: The frequency shift of SBS is highly sensitive to temperature. The system measures the center frequency of the SBS beat frequency component and calculates its drift relative to a reference frequency, for example, the frequency measured at calibration or low power. Since this frequency drift has a clear correlation with fiber temperature, it can be used as the end-face thermal parameter. This method has extremely high measurement sensitivity.

[0165] Finally, the boundary conditions of the turbulence model are updated based on the acquired end-face thermal parameters. Whether the end-face thermal parameters are obtained by measuring the anti-Stokes component power or by measuring the SBS frequency drift, they will be used as real-time updated values ​​and fed back into the turbulence model to dynamically update its end-face boundary conditions, thereby achieving closed-loop adaptive correction of the entire risk assessment model.

[0166] Optionally, the calculation of the second eigenvalue characterizing the temporal statistical properties of the backscattered light includes:

[0167] The backscattered light corresponding to the second real-time parameter sequence is sent into a photodetector to generate an electrical signal corresponding to the intensity of the backscattered light.

[0168] The amplitude of the electrical signal is sampled within a preset time window to obtain an amplitude sample sequence;

[0169] Calculate the second and fourth central moments of the amplitude sample sequence;

[0170] Based on the second-order central moment and the fourth-order central moment, the kurtosis value of the electrical signal is calculated, and the kurtosis value is used as the second feature value.

[0171] Research has found that instantaneous power spikes caused by the random interference of chaotic pump light and end-face reflected light are low-probability, high-amplitude burst events. Traditional monitoring methods based on signal average power or variance are not sensitive to the early stages of such events because sporadic spikes contribute little to the overall average power of the signal and are easily drowned out by background noise, leading to a lag in risk warnings. The essence of the problem is that these risk events alter the probability distribution of signal amplitude, rather than its average value.

[0172] This embodiment introduces advanced statistical analysis methods, particularly calculating the kurtosis of the signal, to directly quantify the morphological changes of the signal probability density function. Kurtosis is an indicator that is extremely sensitive to the thickness of the "tail" of a probability distribution. Its advantage is that it can detect the occurrence of dangerous instantaneous spike events through a sharp increase in the kurtosis value, even when the average power of the signal has not changed significantly, thereby achieving earlier and more sensitive risk warnings than traditional methods.

[0173] In specific implementation, the backscattered light acquired in step S101, characterized by the second real-time parameter sequence, is sent to a high-speed photodetector to convert the instantaneous change in its light intensity into a continuous analog electrical signal. Subsequently, this electrical signal is passed through a high-speed analog-to-digital converter and sampled at a high sampling rate within each preset analysis period, for example, a very short time window, thereby obtaining a string of discrete digital values. This string of digital values ​​constitutes the amplitude sample sequence referred to in this step.

[0174] Next, the second and fourth central moments of the amplitude sample sequence are calculated. This has already been explained above and will not be repeated here.

[0175] Then, the kurtosis value is calculated based on the central moments. According to a pre-defined statistical formula, the second and fourth central moments obtained in the previous step are combined to calculate the kurtosis value. For example, the fourth central moment can be divided by the square of the second central moment to obtain a normalized kurtosis index. This kurtosis value can very sensitively quantify the degree of "fat tail" in the probability distribution of signal amplitude. Background noise with a near-Gaussian distribution will have a kurtosis value that stabilizes at a low baseline level. However, when instantaneous power spikes caused by random interference occur, even if the probability of these spikes is low, the kurtosis value will increase sharply and significantly.

[0176] Finally, this calculated kurtosis value is output as the second feature value characterizing the current random interference risk intensity. This second feature value will be sent to the risk assessment module in step S104 and fused with the first feature value to provide a key dimension characterizing the instantaneous risk at the end face.

[0177] Furthermore, this paper considers how to precisely separate the end-face reflection random interference signal, which is a direct precursor to damage, from other types of background noise, such as Rayleigh scattering noise, pump source intensity noise, and detector noise, with high specificity. The aforementioned kurtosis-based method focuses on the overall shape of the signal probability distribution, while this embodiment aims to achieve "targeted identification" of specific risk sources by utilizing the physical determinism of events.

[0178] This embodiment leverages a key physical insight: the random interference caused by reflections from the fiber optic end face. Although the intensity of this interference fluctuates randomly, its location is physically fixed. This implies that the reflected signal, relative to the monitoring point, must have a definite and predictable round-trip time delay. This embodiment introduces autocorrelation function analysis, a time structure analysis tool, to accurately search for this inherent time delay fingerprint within complex time-domain signals.

[0179] In this way, the risk signal strength from this specific location on the fiber end face can be directly located and quantified, greatly eliminating interference from other irrelevant noise, thus achieving extremely high detection specificity and reliability.

[0180] In this specific implementation, the round-trip time delay of the backscattered light generated at a monitoring point by the fiber end face is predetermined. Before execution of this embodiment, the round-trip time delay needs to be predetermined. This round-trip time delay is a fixed value related to the amplifier structure; it depends on the fiber length from the monitoring point, for example, the location of the second fiber coupler described in step S101, to the far end face of the gain fiber, and the propagation speed of light in the fiber core. This value can be directly calculated using design parameters, or, more precisely, measured during equipment manufacturing or calibration using standard testing instruments such as an optical time-domain reflectometer, and stored as a constant parameter in the memory of the monitoring and protection system.

[0181] Next, the backscattered light is converted into an electrical signal. As described in the previous embodiment, the backscattered light obtained in step S101, characterized by the second real-time parameter sequence, is sent to a high-speed photodetector and converted from analog to digital to generate a series of discrete digital electrical signal sample sequences.

[0182] Next, the autocorrelation function of the electrical signal is calculated within a preset time window. After obtaining the digital sampling sequence of the electrical signal, the autocorrelation function is calculated for the sequence within each preset analysis period. The core of this calculation is to perform a series of delay, multiplication, and accumulation operations on the signal sequence and itself. Specifically, the entire sequence is shifted by different time lengths, and then the corresponding points of the original sequence and the shifted sequence are multiplied point by point, and finally all products are added together. By traversing and calculating a series of different shift amounts, an autocorrelation function curve is finally obtained, with time delay as the horizontal axis and correlation magnitude as the vertical axis.

[0183] Finally, the peak value at a specific delay location is extracted as the second feature value. After obtaining the autocorrelation function curve, the entire curve is not analyzed; instead, the focus is precisely on a key location—the region where the time delay is equal to or near the round-trip time delay predetermined in the first step. The peak value at this location is physically derived from the coherence between the light reflected from the fiber endface and the original light. Therefore, the magnitude of this peak value directly and specifically characterizes the intensity of the endface reflection and the severity of the random interference it induces. The processing module can search for a local maximum within a small window near this specific delay location and use this maximum value as the peak amplitude. Alternatively, the area under the curve in the region near the peak value can be calculated as the integrated energy. Finally, this extracted peak amplitude or integrated energy is used as the second feature value in this embodiment.

[0184] The second eigenvalue will then be fed into step S104, providing a key and highly reliable input dimension for the system's comprehensive decision-making, characterizing the risk of end-face reflection interference.

[0185] Optionally, the absolute amplitude of the target correlation peak extracted above will vary with the overall operating power of the Raman amplifier. Even if the physical state (reflectivity) of the fiber end face remains unchanged, the absolute value of the target correlation peak will increase accordingly when the pump power increases. This makes it extremely difficult to set a fixed damage judgment threshold for this peak that is applicable at all power levels, which can easily lead to missed detection at low power and false detection at high power.

[0186] This embodiment utilizes another peak with a clear physical meaning in the autocorrelation function, namely the main correlation peak, as a real-time, internal reference benchmark. Because the main correlation peak is proportional to the total backscattered power, it indirectly reflects the current pump power level. By calculating the ratio of the risk signal peak to the power reference peak, a dimensionless risk index dynamically normalized to the current power level can be obtained.

[0187] This allows for the removal of interference from pump power fluctuations, providing a purer and more direct reflection of abnormal changes in the physical state of the fiber end face. Consequently, subsequent risk assessment thresholds can be set more universally and reliably, greatly enhancing the robustness of the entire protection system.

[0188] In this implementation, the main correlation peak is used as a reference benchmark. After calculating the autocorrelation function curve, this embodiment first analyzes the curve to extract a reference benchmark. According to the mathematical and physical properties of the autocorrelation function, it must have a maximum value at which the time delay is zero, namely the main correlation peak. The amplitude of this main correlation peak is physically proportional to the total average power of the analyzed signal. In the scenario of this application, the signal is backscattered light, and its total power under normal operating conditions is mainly contributed by Rayleigh scattering, which is linearly related to the pump light power. Therefore, this main correlation peak can be used as a real-time, internal reference quantity to characterize the current total backscattered light power level of the amplifier.

[0189] Simultaneously, on the same autocorrelation function curve, at points where the time delay is equal to or near a predetermined round-trip time delay, the target correlation peak is extracted. As mentioned earlier, the amplitude of this target correlation peak characterizes the power of the coherent interference component caused by reflection from the fiber endface at this specific physical location, and is a risk signal that needs to be closely monitored.

[0190] After obtaining the main correlation peak and the target correlation peak, a normalization calculation is performed. Specifically, the amplitude of the target correlation peak is divided by the amplitude of the main correlation peak to obtain a ratio. This ratio is the coherent / incoherent scattering ratio. Since it is a relative quantity, it largely eliminates the influence of proportional changes in the two peaks caused by fluctuations in pump power itself.

[0191] A healthy, reflectivity-stable end facet will maintain a relatively stable, low ratio regardless of whether it operates at low or high power. Only when the physical state of the end facet deteriorates, such as with contamination or microcracks, leading to an abnormally high reflectivity, will the target correlation peak increase disproportionately faster than the main correlation peak, resulting in a significant and easily detectable jump in the ratio. Therefore, this embodiment uses this highly stable and physically meaningful ratio, which has undergone dynamic adaptive normalization, as the final second eigenvalue.

[0192] This second characteristic value will be fed into step S104. Because it has been intrinsically calibrated for the effects of power fluctuations, it can provide a more stable and reliable input for subsequent risk assessment and decision-making, thereby allowing the system to adopt more universal triggering conditions.

[0193] Furthermore, the risk of end-face damage in high-power Raman amplifiers is not simply a linear superposition of the two risk sources: spectral inhomogeneity (the first eigenvalue) and end-face reflection interference (the second eigenvalue). The coupling effect and relative importance of this risk are closely related to the amplifier's macroscopic operating state, especially its power level. Using fixed, static weighting coefficients for weighted combinations cannot accurately reflect this dynamic, nonlinear risk relationship, potentially leading to oversensitivity and false alarms in the low-power region, or undersensitivity and missed alarms in the high-power region.

[0194] In this embodiment, the weighting coefficients of the risk assessment model should not be static but rather a function of the current amplifier operating state. By monitoring the amplifier's macroscopic operating parameters in real time, such as pump power, and dynamically adjusting the weights used to combine the two independent eigenvalues, the final generated risk index can more realistically simulate the nonlinear process of physical damage. This improves the accuracy and robustness of risk assessment across the entire operating power range, making the basis for protection decisions more reliable.

[0195] In practical implementation, to achieve adaptive adjustment of risk assessment, it is necessary to first obtain a parameter that characterizes the current macroscopic operating state of the amplifier. The most direct and effective parameter is the current average pump power. This parameter can be obtained by averaging the first real-time parameter sequence characterizing the instantaneous power of the pump light obtained in step S101 over a slightly longer time window. Alternatively, it can be indirectly obtained by monitoring the average drive current of the pump source drive circuit and determining its correspondence with the optical power. This average power value reflects the current energy injection level of the amplifier.

[0196] After obtaining the current average pump power, the processing module dynamically determines the weighting coefficients for combining the first and second eigenvalues ​​based on this power value and a preset weighting mapping relationship. This mapping relationship can be implemented as a lookup table. The lookup table stores multiple pump power intervals, and the first and second weighting values ​​corresponding to each interval.

[0197] A core configuration of this weighted mapping relationship is that it is designed so that the second weight value, corresponding to the second eigenvalue characterizing the risk of end-face reflection interference, increases non-linearly with increasing average pump power. This non-linear configuration profoundly reflects the physical mechanism of damage: at low power, the risk of end-face reflection is relatively controllable; however, at high power, the risk of damage from the same reflection intensity increases sharply and disproportionately. Through this dynamic adjustment, the risk assessment model automatically exhibits greater vigilance towards changes in the end-face state in the high-power region.

[0198] After determining the first and second weight values ​​that match the current power level, the processing module performs a weighted combination operation. Specifically, the first feature value obtained in step S102 is multiplied by the dynamically determined first weight value to obtain the first weighted feature value; at the same time, the second feature value obtained in step S103 is multiplied by the dynamically determined second weight value to obtain the second weighted feature value.

[0199] Finally, the two weighted feature values ​​are fused through a preset combination operation to generate the real-time risk indicator. By introducing the current operating state as a dynamic adjustment factor, the real-time risk indicator generated in this embodiment is no longer a static evaluation result, but a higher-dimensional intelligent criterion that can adaptively adjust its internal logic according to the operating conditions. This real-time risk indicator is then sent to step S105. Since this indicator inherently contains a judgment on the risk priority under different operating conditions, it can provide a basis for selecting a more accurate and appropriate protective action.

[0200] Furthermore, traditional protection and response mechanisms typically rely solely on a single aggregated risk value for tiered but singular responses. For example, regardless of the cause, a high risk necessitates a power reduction, which lacks specificity. For instance, abruptly shutting down a recoverable electronic problem would cause unnecessary business interruption; while for the precursors of dangerous physical damage, a mere slight power reduction might not be sufficient to prevent a disaster.

[0201] In making protection decisions, this embodiment not only considers the severity of the total risk, but also traces it back to the independent first characteristic value (spectral risk) and second characteristic value (reflection risk) that constitute the risk, and selects the most appropriate protection action category based on the main source of the risk.

[0202] In this way, the most efficient and appropriate intervention measures can be implemented for different failure modes, thereby maximizing the availability and intelligence of the system while ensuring safety.

[0203] In practical implementation, the triggering conditions for protection decisions are decomposed into sub-conditions corresponding to two independent risk sources. Specifically, at least two independent sub-thresholds are preset: a first threshold for evaluating spectral uniformity and a second threshold for evaluating the risk of end-face reflection interference. The protection module compares the first characteristic value calculated in step S102 with the first threshold; the comparison relationship, such as whether it is greater than the threshold, determines the first sub-condition. Simultaneously, the second characteristic value calculated in step S103 is compared with the second threshold; the comparison relationship determines the second sub-condition.

[0204] Correspondingly, the preset action sets are also divided into different categories according to the protection goals and methods.

[0205] For example, the first type of protective action primarily targets recoverable risks related to the electronic state of the pump source. These actions aim to attempt to restore the system to normal operating condition by adjusting electronic parameters. Specific actions may include: sending a reset command to the chaotic signal generation module of the pump source to attempt to clear spectral degradation that may be caused by temporary software or hardware failures; or sending a specific alarm to the network management center indicating a "pump source health degradation," prompting the need for maintenance and inspection.

[0206] The second type of protective action primarily targets imminent risks that could lead to permanent physical damage. This type of action has the highest priority and aims to reduce the optical power density at the fiber endface as quickly as possible. Specific actions may include: controlling the pump source driver to urgently reduce the output power to a preset safe level; or directly and completely shutting off the current supply to the pump source.

[0207] During runtime, the protection module makes decisions based on the combination of sub-conditions.

[0208] For example, when the system detects that the first characteristic value exceeds the limit while the second characteristic value is normal, that is, the first sub-condition is met but the second sub-condition is not met, the system determines that the current risk mainly comes from the spectrum problem of the pump source, and at this time it will select and execute one or more of the first type of protection actions.

[0209] Once the system detects that the second eigenvalue exceeds the limit, that is, when the second sub-condition is met, regardless of the state of the first eigenvalue, the system determines that there is a direct risk of physical damage to the end face. At this time, it will immediately select and execute one of the second type of protective actions.

[0210] In other embodiments, when two feature values ​​exceed the limit simultaneously, the highest level of combined action, including a second type of protection action and a specific alarm from the first type of protection action, can be executed.

[0211] As mentioned earlier, after selecting the appropriate protective action, the protection module will immediately generate and issue control commands to execute the action, thereby completing a complete and intelligent monitoring and protection closed loop.

[0212] Based on the same inventive concept, this application also provides a high-power Raman amplifier over-excitation damage monitoring and protection system corresponding to the high-power Raman amplifier over-excitation damage monitoring and protection method. Since the principle of the system in this application is similar to the high-power Raman amplifier over-excitation damage monitoring and protection method described above in this application, the implementation of the system can refer to the implementation of the method, and the repeated parts will not be described again.

[0213] Reference Figure 4The diagram shown is a schematic of a high-power Raman amplifier overexcitation damage monitoring and protection system provided in an embodiment of this application. The system includes:

[0214] Acquisition module 10 is used to acquire a first real-time parameter sequence and a second real-time parameter sequence of the Raman amplifier, wherein the first real-time parameter sequence represents the state of the pump light and the second real-time parameter sequence represents the state of the backscattered light.

[0215] The first processing module 20 calculates a first characteristic value characterizing the spectral uniformity of the pump light based on the first real-time parameter sequence; and calculates a second characteristic value characterizing the temporal statistical properties of the backscattered light based on the second real-time parameter sequence, wherein the temporal statistical properties are related to the random interference between the pump light and the light reflected from the fiber end face.

[0216] The second processing module 30 generates a real-time risk index characterizing the current end-face damage risk level by weighting and combining the first feature value and the second feature value.

[0217] The protection module 40 is used to select a protection action corresponding to the trigger condition from an action set including at least two protection actions in response to the real-time risk indicator meeting the preset trigger condition.

[0218] Those skilled in the art will understand that, in the methods described above in the specific embodiments, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic. It should be understood that determining B based on A does not mean determining B solely based on A; B can also be determined based on A and / or other information.

[0219] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

Claims

1. A method for monitoring and protecting high-power Raman amplifiers against overexcitation damage, characterized in that, include: Obtain a first real-time parameter sequence and a second real-time parameter sequence of the Raman amplifier, wherein the first real-time parameter sequence represents the state of the pump light and the second real-time parameter sequence represents the state of the backscattered light; Based on the first real-time parameter sequence, a first characteristic value characterizing the uniformity of the pump light spectrum is calculated; Based on the second real-time parameter sequence, a second characteristic value characterizing the time-domain statistical properties of the backscattered light is calculated, the time-domain statistical properties being related to the random interference between the pump light and the optical fiber end-face reflected light; A real-time risk index representing the current end-face damage risk level is generated by weighting and combining the first feature value and the second feature value. In response to the real-time risk indicator meeting the preset trigger condition, a protective action corresponding to the trigger condition is selected from a set of actions including at least two protective actions.

2. The method for monitoring and protecting against overexcitation damage to high-power Raman amplifiers according to claim 1, characterized in that, The first characteristic value used to characterize the uniformity of the pump light spectrum includes: The first real-time parameter sequence is modeled as a corresponding optical fluid velocity field; The power spectral density of the pump light spectrum is obtained by solving the evolution of the photofluid velocity field based on the turbulence model. A first characteristic value characterizing spectral uniformity is determined based on the power spectral density, wherein the first characteristic value is a turbulent kinetic energy dissipation index calculated based on the power spectral density.

3. The method for monitoring and protecting against overexcitation damage to high-power Raman amplifiers according to claim 2, characterized in that, The calculation of the first characteristic value characterizing the uniformity of the pump light spectrum also includes: Monitor the thermal parameters of the optical fiber end face to determine the end face boundary conditions of the photofluid velocity field; Based on the changes in the end-face boundary conditions, the solution process of the turbulence model is modified, or the calculated turbulent kinetic energy dissipation index is corrected.

4. The method for monitoring and protecting against overexcitation damage to high-power Raman amplifiers according to claim 3, characterized in that, The end-face boundary conditions for determining the photofluid velocity field include: An optical path coupling device with a preset splitting ratio is used to separate the local oscillating light from the forward-propagating pump light, and the local oscillating light is coherently mixed with the backscattered light to generate a mixed beam. The mixed beam is fed into a balanced photodetector or a single-ended photodetector to generate an electrical signal characterizing the beat frequency information between the local oscillating light and the backscattered light. The electrical signal is subjected to spectral analysis to identify and measure the power of the beat frequency component corresponding to anti-Stokes Raman scattering, or the center frequency drift of the beat frequency component corresponding to stimulated Brillouin scattering, and the power or the center frequency drift is used as the end face thermal parameter. The end-face boundary conditions of the turbulence model are updated based on the end-face thermal parameters.

5. The method for monitoring and protecting against overexcitation damage to high-power Raman amplifiers according to claim 1, characterized in that, The second eigenvalue, which characterizes the time-domain statistical properties of the backscattered light, includes: The backscattered light corresponding to the second real-time parameter sequence is sent into a photodetector to generate an electrical signal corresponding to the intensity of the backscattered light. The amplitude of the electrical signal is sampled within a preset time window to obtain an amplitude sample sequence; Calculate the second and fourth central moments of the amplitude sample sequence; Based on the second-order central moment and the fourth-order central moment, the kurtosis value of the electrical signal is calculated, and the kurtosis value is used as the second feature value.

6. The method for monitoring and protecting against overexcitation damage to high-power Raman amplifiers according to claim 1, characterized in that, The second eigenvalue, which characterizes the time-domain statistical properties of the backscattered light, further includes: The round-trip time delay of the backscattered light generated at a monitoring point by the fiber end face is predetermined; The backscattered light corresponding to the second real-time parameter sequence is sent into a photodetector to generate an electrical signal; The autocorrelation function of the electrical signal is calculated within a preset time window to obtain the autocorrelation function curve; Extract the peak amplitude or integral energy on the autocorrelation function curve at a time delay equal to or close to the round-trip time delay, and use the peak amplitude or integral energy as the second characteristic value.

7. The method for monitoring and protecting against overexcitation damage to high-power Raman amplifiers according to claim 6, characterized in that, The step of extracting the peak amplitude or integral energy on the autocorrelation function curve at a time delay equal to or near the round-trip time delay, and using the peak amplitude or integral energy as the second feature value, includes: Obtain the main correlation peak value on the autocorrelation function curve at the point where the time delay is zero, where the main correlation peak value represents the total backscattered light power; Obtain the target correlation peak on the autocorrelation function curve at a time delay equal to or close to the round-trip time delay, wherein the target correlation peak represents the power of the coherent interference component caused by end-face reflection; Calculate the ratio of the target correlation peak value to the main correlation peak value; And the ratio is used as the second characteristic value.

8. The method for monitoring and protecting against overexcitation damage to a high-power Raman amplifier according to claim 5 or 6, characterized in that, The step of generating a real-time risk indicator characterizing the current end-face damage risk level by weighting and combining the first and second feature values ​​includes: The current average pump power of the Raman amplifier is obtained in real time; Based on the current average pump power, a first weight value and a second weight value corresponding to the current average pump power are determined from a preset weight mapping relationship, wherein the weight mapping relationship is configured such that the second weight value increases non-linearly with the increase of the current average pump power. Multiplying the first feature value by the first weight value yields the first weighted feature value, and multiplying the second feature value by the second weight value yields the second weighted feature value; The first weighted feature value and the second weighted feature value are combined to generate the real-time risk indicator.

9. The method for monitoring and protecting against overexcitation damage to a high-power Raman amplifier according to claim 8, characterized in that, The step of selecting a protective action corresponding to the trigger condition from a set of actions including at least two protective actions in response to the real-time risk indicator meeting a preset trigger condition includes: The triggering conditions of the real-time risk indicator are decomposed into a first sub-condition determined based on the comparison relationship between the first feature value and the first threshold, and a second sub-condition determined based on the comparison relationship between the second feature value and the second threshold. The set of actions includes a first type of protective action and a second type of protective action; When the first sub-condition is met but the second sub-condition is not met, the first type of protection action is selected. The first type of protection action is designed to adjust the electronic parameters related to the pump light spectrum. When the second sub-condition is met, the second type of protection action is selected, which includes reducing or shutting down the power of the pump source.

10. A high-power Raman amplifier over-excitation damage monitoring and protection system, characterized in that, include: The acquisition module is used to acquire a first real-time parameter sequence and a second real-time parameter sequence of the Raman amplifier, wherein the first real-time parameter sequence represents the state of the pump light and the second real-time parameter sequence represents the state of the backscattered light. The first processing module calculates a first characteristic value characterizing the uniformity of the pump light spectrum based on the first real-time parameter sequence. Based on the second real-time parameter sequence, a second characteristic value characterizing the time-domain statistical properties of the backscattered light is calculated, the time-domain statistical properties being related to the random interference between the pump light and the optical fiber end-face reflected light; The second processing module performs a weighted combination of the first feature value and the second feature value to generate a real-time risk indicator that characterizes the current end face damage risk level. The protection module is used to select a protection action corresponding to the trigger condition from a set of actions including at least two protection actions in response to the real-time risk indicator meeting the preset trigger condition.

Images