Fiber bragg grating production control system and method

Abstract

The invention discloses a fiber bragg grating production control system and method, and belongs to the technical field of optical fiber sensor manufacturing, and the method comprises the steps: obtaining a pre-stress spectrum curve and an ultraviolet laser parameter set of a target optical fiber before writing, and constructing a feature vector reflecting the disturbance sensitivity of the optical fiber; in the laser writing process, acoustic emission signals and laser parameters are collected in real time, and a joint disturbance function is constructed; identifying a nonlinear stress disturbance event through dynamic matching with the sensitive feature vector, and extracting occurrence time and disturbance intensity of the nonlinear stress disturbance event; constructing a prediction model based on disturbance history, predicting a future disturbance trend and adjusting laser parameters; after writing, a grating reflection spectrum is collected and compared with a target template spectrum, and if the similarity requirement is not met, a disturbance error path is traced back, and parameter compensation writing is executed; according to the invention, real-time perception and response regulation and control of micro-disturbance can be realized, and the writing precision and consistency of the fiber bragg grating are improved.

Description

Technical Field

[0001] This invention relates to the field of fiber optic sensor manufacturing technology, specifically to a fiber optic grating production control system and method. Background Technology

[0002] Fiber Bragg gratings (FBGs), as highly sensitive and stable fiber optic sensing elements, are widely used in aerospace, civil engineering, and intelligent manufacturing. However, the fabrication process of FBGs is extremely sensitive to the microscopic state of the optical fiber, especially during ultraviolet laser irradiation. Microscale nonlinear stress disturbances caused by thermal stress fields, mechanical vibrations, or micro-defects in the substrate can easily lead to grating period instability, reflection peak drift, or even breakage and failure.

[0003] Currently, most fiber Bragg grating fabrication processes employ fixed parameter control or post-processing detection and correction methods, lacking the ability to identify and adjust disturbance sources in real time. In particular, when the stress disturbance intensity is small, the frequency is irregular, and it exhibits non-Gaussian distribution characteristics, traditional control strategies are difficult to detect, leading to the accumulation of latent defects during the production process, forming "latent cracks," which can easily trigger catastrophic failures in subsequent applications.

[0004] Especially in the grating writing process of high-stress optical fibers (such as those used in nuclear power and deep-sea sensing), the periodic non-uniformity caused by such micro-perturbations will significantly reduce the signal-to-noise ratio and accuracy of the reflection spectrum, becoming a key bottleneck restricting the improvement of fiber Bragg grating performance. Summary of the Invention

[0005] The purpose of this invention is to provide a fiber Bragg grating production control system and method to address the shortcomings in the prior art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a fiber Bragg grating production control method, comprising: S100, Obtain the prestress spectrum curve P0 of the target optical fiber before writing, and the selected set of ultraviolet laser parameters L; S200. Construct the disturbance-sensitive feature vector D0 of the optical fiber segment based on the prestressed spectrum curve P0, including multi-scale stress amplitude gradient, frequency distribution entropy and spectral phase disturbance index. S300. During the ultraviolet laser writing process, the acoustic emission signal S(t) and laser incident parameter L(t) of the fiber segment are collected in real time, and their joint perturbation function F(t)=f(S(t),L(t)) within the time window Δt is calculated. S400. Dynamically match the joint perturbation function F(t) with the obtained perturbation sensitive feature vector D0 to determine whether a potential nonlinear stress perturbation event Ek occurs. If a potential nonlinear stress perturbation event Ek occurs, mark its corresponding time tk and perturbation intensity Rk. S500. Based on the corresponding time tk and disturbance intensity Rk, construct a disturbance prediction model M to predict the disturbance trend curve R'(t) in the subsequent time period Δt', and dynamically adjust the laser parameters based on the prediction results. S600. After the grating is written, the reflection spectrum is collected and compared with the target template spectrum. If the correlation coefficient between the two is less than the threshold θ, the error accumulation path of the disturbance intensity Rk in steps S400 to S500 is back-analyzed, and the parameter compensation writing process is executed.

[0007] Preferably, the prestress spectrum curve P0 is the axial prestress spectrum of the optical fiber obtained by micro-strain testing, and L includes laser power, wavelength, focusing size and pulse frequency.

[0008] Preferably, the prestressed spectrum curve P0 is divided into 5 equally spaced frequency segments. Within each frequency segment, the amplitude gradient G(k) of the corresponding segment is obtained. G(k) is equal to the difference in amplitude between two consecutive points divided by the difference in frequency. Then, the average value of all results is calculated.

[0009] Preferably, the amplitudes of all sampling points in the prestressed spectrum curve P0 are normalized so that the total sum of amplitudes is 1, denoted as the normalized amplitude sequence P={p1,p2,...,pn}, where n is the total number of sampling points; the frequency distribution entropy is calculated on the normalized amplitude sequence using the Shannon entropy formula.

[0010] Preferably, the joint perturbation function F(t) is calculated as follows: within a set time window Δt, the acoustic emission signal S(t) and the laser parameter L(t) are standardized and combined into a feature matrix U, and its covariance matrix C is calculated. The average value of the absolute values ​​of the off-diagonal elements is extracted as F(t).

[0011] Preferably, determining whether a potential nonlinear stress disturbance event Ek has occurred includes: Let the first threshold T1 be the mean of the joint disturbance deviation ΔF(t) plus 2 standard deviations, and let the second threshold T2 be the mean plus 3 standard deviations. If ΔF(t) is higher than T1 for 3 consecutive time windows, and at least one of them is higher than T2, then it is identified as a nonlinear stress disturbance event Ek.

[0012] Preferably, a disturbance prediction model M is constructed based on the corresponding time tk and disturbance intensity Rk, including: The time series of disturbance intensity Rk is differentially processed and linearly fitted with the principal components in the disturbance-sensitive feature vector D0 to predict the disturbance intensity R′(k+n) at each time in the future Δt′. When the maximum value R′max of the predicted perturbation trend R′(t) is greater than the second threshold T2, and the average slope K′ is greater than the preset slope threshold κ1, the adjustment of the laser parameter set L(t+Δt′) includes: reducing the laser power L1 by 10% and increasing the focused spot diameter L3 by 15% to reduce the energy density per unit area.

[0013] Preferably, the correlation threshold θ is set to 0.98; when the correlation coefficient ρ≥0.98, the writing result is considered qualified; when ρ<0.98, it is determined that there is a quality deviation in the fiber Bragg grating writing, and the backtracking and compensation mechanism needs to be activated.

[0014] Preferably, during the parameter compensation writing process, if the error εk of a certain disturbance event Ek is positive, the compensation writing strategy is to increase the laser power at that position by 5%; if εk is negative, the laser power is reduced by 5% and the pulse frequency is reduced by 10%.

[0015] The present invention also provides a fiber Bragg grating production control system, comprising: The data acquisition module acquires the prestress spectrum curve P0 of the target optical fiber before writing, as well as the selected set of ultraviolet laser parameters L; The disturbance feature modeling module constructs a disturbance-sensitive feature vector D0 for the optical fiber segment based on the prestressed spectral curve P0, including multi-scale stress amplitude gradient, frequency distribution entropy, and spectral line phase disturbance index. The perturbation function construction module collects the acoustic emission signal S(t) and laser incident parameter L(t) of the fiber segment in real time during the ultraviolet laser writing process, and calculates their joint perturbation function F(t)=f(S(t),L(t)) within the time window Δt. The event determination module dynamically matches the joint perturbation function F(t) with the obtained perturbation sensitive feature vector D0 to determine whether a potential nonlinear stress perturbation event Ek has occurred. If a potential nonlinear stress perturbation event Ek has occurred, its corresponding time tk and perturbation intensity Rk are marked. The disturbance prediction and adjustment module constructs a disturbance prediction model M based on the corresponding time tk and disturbance intensity Rk, which is used to predict the disturbance trend curve R'(t) in the subsequent time period Δt', and dynamically adjusts the laser parameters based on the prediction results; After the grating is written, the compensation control module collects the reflection spectrum and compares it with the target template spectrum. If the correlation coefficient between the two is less than the threshold θ, it backtracks to analyze the error accumulation path of the disturbance intensity Rk in the event judgment module to the disturbance prediction and adjustment module, and executes the parameter compensation writing process.

[0016] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention constructs a perturbation-sensitive vector D0 based on prestress spectrum characteristics and acquires the acoustic emission signal S(t) and laser parameters L(t) in real time during laser writing to construct a joint perturbation function F(t). This enables high-precision identification and quantitative analysis of microscale nonlinear stress perturbation events. This method overcomes the limitations of traditional processes that rely on fixed parameters or post-process detection, possessing millisecond-level time-domain sensing capabilities and effectively identifying high-frequency perturbation behavior, providing real-time assurance for the stability of fiber Bragg grating writing.

[0017] 2. This invention constructs a perturbation prediction model M with structural features involved to achieve dynamic prediction of subsequent perturbation trends and feedforward control of laser parameters. Simultaneously, it combines reflection spectrum quality comparison and error backtracking mechanisms to achieve closed-loop optimization control of the writing process. This method significantly improves grating writing accuracy and consistency, and reduces the risk of performance drift caused by micro-perturbations. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the method of the present invention.

[0020] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1, please refer to Figure 1 As shown in this embodiment, a fiber Bragg grating production control method includes: S100, obtain the prestress spectrum curve P0 of the target optical fiber before writing, and the selected set of ultraviolet laser parameters L.

[0023] The prestress spectrum curve P0 is the response curve of the stress distribution in the axial direction of the target optical fiber before ultraviolet laser writing, as a function of frequency. It reflects the internal stress state accumulated in the fiber during the early stages of processing, transportation, and fixing, and has a feedforward prediction function for laser disturbance response. The acquisition steps are as follows: Cut a fiber segment corresponding to the area to be written, with a length between 50 mm and 80 mm, selecting the specific value based on the focal length characteristics of the writing device. Ensure the fiber end face is clean, without breaks or contamination.

[0024] A piezoelectric stress exciter with frequency scanning function was used to apply axial micro-vibration excitation to the optical fiber. The excitation frequency range was set to 1 Hz to 10 kHz, with a step interval of 5 Hz to ensure coverage of the natural stress response frequency range of each order.

[0025] High-sensitivity laser interferometric displacement sensors are deployed at both ends of the optical fiber to collect the axial vibration amplitude response of the optical fiber and record the response amplitude data at each excitation frequency.

[0026] The collected response data is normalized in amplitude, and a frequency domain spectrum curve is plotted with the excitation frequency as the horizontal axis and the normalized response amplitude as the vertical axis. This spectrum curve is the prestress spectrum curve of the target optical fiber, denoted as P0.

[0027] The ultraviolet laser parameter set L is the set of all laser control parameters used in the grating writing process, which directly determines the stability and spatial distribution characteristics of the interaction between the optical fiber and laser energy.

[0028] Record the current pulse power output value of the ultraviolet laser, in milliwatts, set as L1, typically ranging from 1 milliwatt to 10 milliwatts.

[0029] Determine the center wavelength of the laser to be used, denoted as L2, in nanometers. Typical ultraviolet laser wavelengths are 248 nm (krypton fluoride laser) or 266 nm (frequency-doubled Nd:YAG laser).

[0030] The diameter of the focused spot is measured using a laser focal spot observation instrument, denoted as L3, with the unit being micrometers, typically ranging from 5 micrometers to 20 micrometers.

[0031] The pulse frequency parameter is set to L4, in Hertz, typically ranging from 100 Hertz to 500 Hertz. The pulse frequency directly determines the energy accumulation rate.

[0032] Finally, the ultraviolet laser parameter set L is represented as: L={L1,L2,L3,L4}; where: L1 is the laser power parameter, L2 is the wavelength parameter, L3 is the focused spot diameter, and L4 is the pulse frequency.

[0033] S200. Construct the disturbance-sensitive feature vector D0 of the optical fiber segment based on the prestressed spectrum curve P0, including multi-scale stress amplitude gradient, frequency distribution entropy and spectral phase disturbance index.

[0034] Multi-scale stress amplitude gradient is used to describe the variation trend of fiber prestress response at different frequency scales, reflecting the local inhomogeneity and sensitive range of fiber structure under multi-frequency excitation.

[0035] The frequency axis of the prestressed spectrum curve P0 is divided into 5 equally spaced frequency intervals, namely the first frequency band to the fifth frequency band, corresponding to a frequency range of 1 Hz to 10 kHz, with each segment having a width of 2 kHz.

[0036] Within each frequency band, the average amplitude gradient of adjacent sampling points within that band is calculated. Let there be N sampling points in a frequency band, with the amplitude of the i-th point being V(i) and the frequency being f(i). The amplitude gradient of that band is denoted as G(k), calculated as follows: G(k) equals the difference in amplitude between two consecutive points from the 1st sampling point to the (N-1)th sampling point, divided by the difference in frequency, and then the average of all results is calculated. A total of 5 amplitude gradient values ​​are obtained for each frequency band, forming a multi-scale stress amplitude gradient vector G={G(1),G(2),G(3),G(4),G(5)}.

[0037] Frequency distribution entropy is used to evaluate the complexity and randomness of the prestress spectrum curve in the frequency domain, and measures the degree of dispersion of the fiber stress state from the perspective of information theory.

[0038] The amplitudes of all sampling points in the prestressed spectrum curve P0 are normalized so that the sum of the total amplitudes is 1. This normalized amplitude sequence is denoted as P={p1,p2,...,pn}, where n is the total number of sampling points.

[0039] The frequency distribution entropy H_f of the normalized amplitude sequence is calculated using the Shannon entropy formula. The calculation method is as follows: , where i∈[1,n]; the larger H_f is, the more uniform the prestress distribution of the optical fiber and the less significant peak characteristics it has; the smaller H_f is, the more significant stress concentration region there is, which may lead to local nonlinear disturbances during the writing process.

[0040] The spectral phase perturbation index is used to reflect the non-stationarity characteristics in the prestressed spectrum curve and to identify periodic phase perturbations caused by potential microstructural defects.

[0041] The spectral curve P0 was decomposed into three levels using Daubechies wavelet basis functions to obtain the high-frequency coefficient sets W1, W2 and W3 at each scale.

[0042] In each high-frequency coefficient sequence, the symbol change frequency between adjacent coefficient points is calculated, i.e., the symbol flip rate R(k), and perturbation indices R(1), R(2), and R(3) are obtained at three scales.

[0043] Each R(k) is defined as the number of times adjacent coefficients in a high-frequency sequence differ in sign, divided by the total number of times, representing the severity of the phase change.

[0044] The phase perturbation indices of the three scales are combined to form the spectral line phase perturbation index vector R={R(1),R(2),R(3)}.

[0045] All features extracted from the above three dimensions are combined to form a complete perturbation-sensitive feature vector D0: D0={G(1),G(2),G(3),G(4),G(5),H_f,R(1),R(2),R(3)}; a total of 9 features, which represent the sensitivity of optical fiber to multi-frequency response, information complexity and phase perturbation respectively.

[0046] S300. During the ultraviolet laser writing process, the acoustic emission signal S(t) and laser incident parameter L(t) of the fiber segment are acquired in real time, and their joint perturbation function F(t)=f(S(t),L(t)) within the time window Δt is calculated.

[0047] The acoustic emission signal S(t) is a high-frequency elastic wave signal generated by stress release, microcrack propagation, or local disturbance of the glass substrate during laser writing excitation of the optical fiber. Its changes reflect the dynamic state of the microstructure response. The specific acquisition steps are as follows: Piezoelectric ceramic sensors with a frequency response range of 100 kHz to 1 MHz are adhered to both ends of the target optical fiber segment and coupled and fixed using inorganic colloids to ensure continuous signal transmission.

[0048] Set the sampling frequency of the data acquisition instrument to 5 MHz and the sampling time step to 0.2 microseconds to ensure that all high-frequency disturbance waveforms are captured.

[0049] Each time point t corresponds to an acoustic emission amplitude A(t), in microvolts, and is recorded as an acoustic emission signal sequence S(t)={A(t0),A(t1),A(t2),…,A(tn)}, where t0 to tn are discrete time points within the laser writing cycle.

[0050] The laser incident parameter L(t) represents the real-time changes of various control parameters during the ultraviolet laser writing process and is the direct excitation source of dynamic disturbances. The parameter acquisition method is as follows: The parameters monitored include laser power L1(t), laser wavelength L2(t), focused spot size L3(t), and pulse frequency L4(t), which are consistent with the static parameter L in step S100.

[0051] The parameters are read in real time through a four-channel sensing interface connected to the laser controller. The sampling frequency is set to 5 MHz, the same as S(t), and hardware clock synchronization is performed to ensure data consistency. The laser incident parameter sequence is represented as L(t) = {L1(t), L2(t), L3(t), L4(t)}, and each time t corresponds to a complete set of laser parameter values.

[0052] Define a sliding time window Δt, and calculate the joint perturbation characteristic function F(t) within this window. The specific steps are as follows: Define Δt as 2 milliseconds, and the number of sampling points contained in the window is 10,000. The window is updated by sliding in steps of 0.5 milliseconds.

[0053] The joint perturbation function F(t) is defined as the multidimensional coordinated response characteristic between the acoustic emission signal and the laser parameters within a time window of Δt. The construction process of the function f(S(t),L(t)) includes: The acoustic emission amplitude sequence within the Δt time window is standardized to construct a standardized signal vector S′; the laser parameter sequences within the same time period are respectively min-max normalized to form a standardized laser vector L′.

[0054] Combine S′ and L′ into a joint feature matrix U; calculate the covariance matrix C of the joint feature matrix U, which has a dimension of 5×5, where the first dimension corresponds to the acoustic emission signal and the other four dimensions correspond to the laser parameters; any item C(i,j) in the covariance matrix represents the linear cooperative change trend of the i-th feature and the j-th feature in the time interval Δt.

[0055] The average absolute value of the off-diagonal elements in the covariance matrix is ​​extracted and defined as the joint perturbation strength index F(t); that is, F(t) equals the sum of the absolute values ​​of all elements except the main diagonal divided by the number of elements. The larger this value, the stronger the cooperative perturbation between the laser parameter changes and the acoustic emission signal, and the more likely there is an unsteady change behavior in the microstructure response.

[0056] Each Δt time window corresponds to a joint perturbation function value F(t). During the entire laser writing cycle, F(t) is serialized into a time function curve, which serves as the input basis for perturbation event identification and intensity quantification in the subsequent step S400.

[0057] S400. Dynamically match the joint perturbation function F(t) with the obtained perturbation-sensitive feature vector D0 to determine whether a potential nonlinear stress perturbation event Ek occurs. If a potential nonlinear stress perturbation event Ek occurs, mark its corresponding time tk and perturbation intensity Rk.

[0058] To achieve comparability between the joint perturbation function and the inherent perturbation sensitivity characteristics of optical fibers, a perturbation reference response model is first constructed to describe the expected response relationship between the joint perturbation function F(t) and the perturbation-sensitive eigenvector D0 under normal write conditions. The nine eigenvalues ​​in the perturbation-sensitive eigenvector D0 are arranged in the following order to form a baseline feature sequence: Multiscale stress amplitude gradient characteristics G(1) to G(5); Frequency distribution entropy H_f; Spectral line phase perturbation indices R(1), R(2), and R(3).

[0059] The above nine feature quantities are assigned weight coefficients W(1) to W(9). The weight coefficients are determined as follows: Fifty optical fibers from the same batch were selected as training samples; their joint perturbation function was recorded under conditions of no abnormal writing; the correlation coefficient between each feature and the joint perturbation function was calculated; and the normalized correlation coefficient was used as the weight value of the corresponding feature.

[0060] The perturbation reference function, denoted as F0, is obtained as follows: F0 equals the sum of the products of each feature in the perturbation-sensitive eigenvector D0 and its corresponding weight coefficient. This reference function represents the theoretical joint perturbation response level of the optical fiber to laser writing in the absence of nonlinear stress perturbations.

[0061] During the ultraviolet laser writing process, for each time window Δt, the deviation of the joint perturbation function value F(t) from the perturbation reference function F0 is calculated to characterize the real-time perturbation state.

[0062] Define the joint disturbance deviation ΔF(t), which is calculated as follows: ΔF(t) equals the joint disturbance function value F(t) within the current time window minus the disturbance reference function F0.

[0063] Arrange ΔF(t) corresponding to all time windows during the entire writing process in chronological order to form a joint perturbation deviation time series for subsequent event identification.

[0064] To accurately distinguish between normal fluctuations and potential nonlinear stress disturbances, a dual-threshold judgment rule is introduced to identify the deviation of the joint disturbance. The mean and standard deviation of ΔF(t) in the training samples are statistically analyzed; the first threshold T1 is defined as the mean plus twice the standard deviation; the second threshold T2 is defined as the mean plus three times the standard deviation.

[0065] A potential nonlinear stress disturbance event Ek is determined to have occurred when the following two conditions are met: ΔF(t) is continuously greater than the first threshold T1 for at least three consecutive adjacent time windows; and ΔF(t) is greater than the second threshold T2 for at least one of these time windows. This determination rule is used to eliminate false judgments caused by transient noise and ensure that the identification results correspond to the true nonlinear stress response behavior.

[0066] Once a potential nonlinear stress disturbance event Ek is determined to have occurred, the time of occurrence of the event is marked.

[0067] The time window in which ΔF(t) first exceeds the first threshold T1 is defined as the time when the disturbance event occurs; this time is denoted as tk.

[0068] If ΔF(t) satisfies the event triggering condition again in a subsequent time period, and the time interval between the event and the previous event is greater than 5 times the time window Δt, it is determined to be a new disturbance event, and the number is incremented sequentially.

[0069] The disturbance intensity Rk is used to describe the impact of potential nonlinear stress disturbance events on the stability of fiber writing. Its quantification method is as follows: For the entire time window covered by the disturbance event Ek, the maximum value of ΔF(t) is extracted and denoted as ΔF_max. The disturbance intensity Rk is defined as the ratio of ΔF_max to the second threshold T2.

[0070] Rk being close to 1 indicates that the disturbance has just reached the nonlinear threshold; Rk greater than 1 indicates that the degree of disturbance increases proportionally; The larger the value of Rk, the more pronounced the nonlinear stress release or structural response behavior occurs inside the optical fiber. The output of step S400 includes: a sequence of potential nonlinear stress disturbance events {E1, E2, ..., Ek}; the occurrence time stamp tk for each event; and the disturbance intensity value Rk for each event.

[0071] S500. Based on the corresponding time tk and disturbance intensity Rk, a disturbance prediction model M is constructed to predict the disturbance trend curve R'(t) in the subsequent time period Δt', and the laser parameters are dynamically adjusted based on the prediction results.

[0072] The disturbance prediction model M is a time series regression model built based on the previously identified disturbance event time series {tk} and its corresponding intensity series {Rk}, used to predict the evolution trend of disturbance intensity within the future time period Δt′. The disturbance event series obtained from step S400 includes: the time stamp tk of each disturbance event; the disturbance intensity Rk of each disturbance event; and constitutes the disturbance historical data set: D={(t1,R1),(t2,R2),...,(tk,Rk)}.

[0073] Set the prediction time period Δt′ to 10 milliseconds, divide it into several prediction steps, each prediction step is set to 1 millisecond, and predict a total of 10 future time points, denoted as t(k+1), t(k+2), ..., t(k+10).

[0074] An improved autoregressive moving average model, denoted as M, is used to construct a disturbance trend prediction model. The model form is an autoregressive model with external input (ARX). The prediction form of M is: R′(k+n) equals the sum of the weighted regression term of historical disturbance intensity and the linear projection of the disturbance-sensitive feature vector D0. This model comprehensively considers the mapping relationship between the time series trend of disturbance intensity and the characteristics of the optical fiber itself, thereby improving prediction accuracy.

[0075] The steps for constructing model M are as follows: Step 1: Perform time difference processing on the historical perturbation sequence to obtain the intensity increment ΔRk; Step 2: Use the least squares method to fit the linear relationship between ΔRk and the time interval Δtk to obtain the first-order regression coefficients; Step 3: Project the disturbance-sensitive feature vector D0 onto a set of feature basis vectors most correlated with the disturbance intensity to obtain the structural disturbance tendency factor; Step 4: Integrate the time trend coefficient and the structural tendency factor to construct a multivariate linear prediction function and obtain the disturbance prediction model M.

[0076] After the model is built, the model M is used to predict the intensity of the disturbance in the future time period Δt′, and the disturbance trend curve R′(t) is output.

[0077] The model M is used to calculate the predicted disturbance intensity value R′t(k+n) for each future time point t(k+n) (n∈[1,10]).

[0078] The prediction results form a disturbance trend curve: R′(t)={R′t(k+1),R′t(k+2),...,R′t(k+10)}; this curve represents the trend of disturbance intensity change that may occur in the optical fiber in the future Δt′.

[0079] To provide a quantitative basis for subsequent laser parameter adjustments, the following two key prediction features are extracted from R′(t): the predicted value of the maximum disturbance intensity R′max; and the average slope of the curve K′, which is used to represent the growth trend of the disturbance intensity.

[0080] Based on the feature values ​​extracted from the predicted perturbation trend curve R′(t), targeted adjustments are made to the ultraviolet laser parameter set L(t) to achieve proactive intervention in the perturbation. The ultraviolet laser parameter set L(t) includes the following four items: L1(t): laser power; L2(t): laser wavelength; L3(t): focused spot size; L4(t): pulse frequency. The adjustment strategy is determined based on the values ​​of R′max and K′: if R′max ≥ the second threshold T2 and K′ ≥ the positive threshold κ1, the perturbation is considered to be intensifying; if R′max is between the first threshold T1 and the second threshold T2, and K′ is close to 0, the perturbation is considered to be in a critical state; if R′max < the first threshold T1, the current laser parameters are maintained unchanged.

[0081] When the predicted disturbance is an escalating trend, the following dynamic adjustment operations are performed on the laser parameters: the laser power L1(t+Δt′) is reduced by 10%; the focused spot size L3(t+Δt′) is increased by 15% to reduce the energy density per unit area; the wavelength L2 and pulse frequency L4 are kept constant to avoid introducing new variables that could cause interference. These parameter adjustments constitute a limited-amplitude disturbance suppression strategy to prevent systemic instability.

[0082] S600. After the grating is written, the reflection spectrum is collected and compared with the target template spectrum. If the correlation coefficient between the two is less than the threshold θ, the error accumulation path of the disturbance intensity Rk in steps S400 to S500 is back-analyzed, and the parameter compensation writing process is executed.

[0083] After the fiber grating is written, the optical fiber reflection performance is first measured to obtain the actual reflection spectrum R_f of the current grating.

[0084] Measurements were performed using a high-resolution spectrometer with a spectral range of 1520 nm to 1580 nm and a resolution better than 0.01 nm.

[0085] The reflectance spectrum R_f is represented as a set of functional relationships between wavelength and reflectance, with a sampling interval of 0.01 nanometers, denoted as: R_f(λ)={(λ1,r1),(λ2,r2),...,(λn,rn)}; where λ represents wavelength, r represents reflectance, and n is the number of sampling points.

[0086] The measured reflection spectrum R_f is compared point by point with the target template spectrum Rm, and the Pearson correlation coefficient ρ between the two is calculated as a quantitative indicator of the degree of consistency of reflection performance.

[0087] The target template spectrum Rm is the ideal reflection spectrum response designed under the current process conditions. Its wavelength range is consistent with R_f, and its form is: Rm(λ)={(λ1,r1′),(λ2,r2′),...,(λn,rn′)}; The Pearson correlation coefficient ρ between the reflectance sequences {r1,r2,...,rn} and {r1′,r2′,...,rn′} is calculated using the formula: ρ equals the covariance of the measured and template reflectances divided by the product of their standard deviations. The value ranges from -1 to 1, with a value closer to 1 indicating a better match.

[0088] To determine whether the grating reflection spectrum meets the preset quality standard, an evaluation threshold θ is set: The correlation threshold θ is set to 0.98; When the correlation coefficient ρ ≥ 0.98, the writing result is considered acceptable; If ρ < 0.98, it is determined that there is a quality deviation in the fiber Bragg grating writing, and a backtracking and compensation mechanism needs to be activated.

[0089] After detecting a quality deviation, the intensity deviation generated during the disturbance identification and prediction process in steps S400 to S500 is analyzed back, and the error accumulation path is extracted to locate the compensation target.

[0090] Extract the set of disturbance events {E1, E2, ..., Ek} and their corresponding times tk and predicted disturbance intensity R′(tk) from step S400. Compare the predicted disturbance intensity R′(tk) with the actual measured disturbance intensity Rk and calculate the error term εk: Construct the error sequence E={ε1,ε2,...,εk}.

[0091] Based on the error magnitude and event duration, calculate the impact weight of each disturbance event on the overall write bias: Wk = |εk| × τk; where τk is the duration of the disturbance event Ek. ​​Sort all events according to Wk, and select the top m events with the greatest impact as key compensation targets (m is usually set to 3 to 5).

[0092] Based on the error accumulation path analysis results, a laser parameter compensation rewriting operation is performed on the written grating to correct local reflection performance deviations.

[0093] The spatial locations corresponding to the m disturbance events with the largest positioning error weights are determined based on the mapping relationship between fiber writing speed and tk, and the corresponding fiber segment length range is determined.

[0094] For each target location, the following compensation strategy is selected based on the sign and magnitude of the error direction εk: If εk is positive (indicating underprediction), then increase the local laser power L1(t) by 5%; If εk is negative (over-prediction), then reduce the power by 5% and simultaneously reduce the pulse frequency L4(t) by 10%. The spot size L3 remains unchanged to avoid introducing spot reconstruction errors; The wavelength L2 is fixed and locked to the main reflection peak wavelength ±0.1 nm during the compensation write.

[0095] The compensation writing adopts a scan rewrite mode, in which the laser beam scans the target position 1 to 3 times at a constant speed, and the number of cycles is set according to the absolute value of the error.

[0096] Example 2, please refer to Figure 2 As shown in this embodiment, a fiber Bragg grating production control system includes: The data acquisition module acquires the prestress spectrum curve P0 of the target optical fiber before writing, as well as the selected set of ultraviolet laser parameters L; The disturbance feature modeling module constructs a disturbance-sensitive feature vector D0 for the optical fiber segment based on the prestressed spectral curve P0, including multi-scale stress amplitude gradient, frequency distribution entropy, and spectral line phase disturbance index. The perturbation function construction module collects the acoustic emission signal S(t) and laser incident parameter L(t) of the fiber segment in real time during the ultraviolet laser writing process, and calculates their joint perturbation function F(t)=f(S(t),L(t)) within the time window Δt. The event determination module dynamically matches the joint perturbation function F(t) with the obtained perturbation sensitive feature vector D0 to determine whether a potential nonlinear stress perturbation event Ek has occurred. If a potential nonlinear stress perturbation event Ek has occurred, its corresponding time tk and perturbation intensity Rk are marked. The disturbance prediction and adjustment module constructs a disturbance prediction model M based on the corresponding time tk and disturbance intensity Rk, which is used to predict the disturbance trend curve R'(t) in the subsequent time period Δt', and dynamically adjusts the laser parameters based on the prediction results; After the grating is written, the compensation control module collects the reflection spectrum and compares it with the target template spectrum. If the correlation coefficient between the two is less than the threshold θ, it backtracks to analyze the error accumulation path of the disturbance intensity Rk in the event judgment module to the disturbance prediction and adjustment module, and executes the parameter compensation writing process.

[0097] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method of controlling production of fiber gratings, characterized by: Comprising: S100, obtaining the pre-stress spectrum curve P0 of the target optical fiber before writing, and a selected set of ultraviolet laser parameters L; S200, constructing a perturbation sensitive feature vector D0 of the fiber segment according to the pre-stress spectrum curve P0, including multi-scale stress amplitude gradient, frequency distribution entropy and spectrum line phase perturbation index; S300, during the ultraviolet laser writing process, real-time acquisition of the acoustic emission signal S(t) and the laser incident parameter L(t) of the fiber segment, and calculation of the joint perturbation function F(t)=f(S(t), L(t)) within the time window Δt; S400, dynamic matching of the joint perturbation function F(t) with the obtained perturbation sensitive feature vector D0, judgment of whether a potential nonlinear stress perturbation event Ek occurs, if there is a potential nonlinear stress perturbation event Ek, marking the corresponding time tk and the perturbation intensity Rk; S500, constructing a perturbation prediction model M according to the corresponding time tk and the perturbation intensity Rk, to predict the perturbation trend curve R'(t) within the subsequent Δt' time period, and dynamically adjusting the laser parameters based on the prediction result; S600, after the grating writing is completed, the reflection spectrum is collected and compared with the target template spectrum; if the correlation coefficient of the two is less than the threshold θ, the error accumulation path of the perturbation intensity Rk in steps S400 to S500 is analyzed in retrospect, and the parameter compensation writing process is performed.

2. A fiber grating production control method according to claim 1, characterized by: Wherein, The pre-stress spectrum curve P0 is the axial pre-stress spectrum of the optical fiber obtained by micro-strain test, and L includes laser power, wavelength, focusing size and pulse frequency.

3. The method of claim 1, wherein: the fiber Bragg grating is formed in a fiber having a core with a core diameter of 50 μm or less. The pre-stress spectrum curve P0 is divided into 5 equal interval frequency segments, in each frequency segment, the amplitude gradient G(k) of the corresponding segment is obtained, G(k) is equal to the difference between the amplitudes of two consecutive points divided by the difference between the frequencies, and then the average value of all results is obtained.

4. The method of claim 1, wherein: The amplitudes of all sampling points in the pre-stress spectrum curve P0 are normalized to make the total amplitude sum equal to 1, and the normalized amplitude sequence P={p1, p2,..., pn} is recorded, where n is the total number of sampling points; the frequency distribution entropy is calculated by using the Shannon entropy formula on the normalized amplitude sequence.

5. The method of claim 1, wherein: The calculation method of the joint perturbation function F(t) is: within the set time window Δt, the acoustic emission signal S(t) and the laser parameter L(t) are standardized, combined into a feature matrix U, and the covariance matrix C is calculated, and the average value of the absolute values of the non-diagonal elements is taken as F(t).

6. The method of claim 1, wherein: Judgment of whether a potential nonlinear stress perturbation event Ek occurs, including: Set the first threshold T1 as the mean value of the joint perturbation deviation ΔF(t) plus 2 times the standard deviation, and set the second threshold T2 as the mean value plus 3 times the standard deviation, if ΔF(t) is higher than T1 for 3 consecutive time windows, and at least one is higher than T2, it is identified as a nonlinear stress perturbation event Ek.

7. The method of claim 1, wherein: Construction of the perturbation prediction model M according to the corresponding time tk and the perturbation intensity Rk, including: ​ Difference processing of the perturbation intensity Rk time series, linear fitting combined with the principal components in the perturbation sensitive feature vector D0 to predict the perturbation intensity R'(k+n) at each time within the future Δt'; When the maximum value R'max of the predicted disturbance trend R'(t) is greater than the second threshold T2, and the average slope K' is greater than the preset slope threshold κ1, the adjustment of the laser parameter set L(t+Δt') includes: reducing the laser power L1 by 10%, and increasing the focused spot diameter L3 by 15%, so as to reduce the energy density per unit area.

8. The method of claim 1, wherein: The correlation threshold θ is set to 0.98; when the correlation coefficient ρ≥0.98, it is considered that the writing result is qualified; when ρ<0.98, it is considered that there is a quality deviation in the fiber grating writing, and the backtracking and compensation mechanism needs to be started. ​ 9. A method of controlling the production of fibre optical gratings according to claim 8, characterised in that: In the parameter compensation writing process, if the error εk of a disturbance event Ek is positive, the compensation writing strategy is to increase the laser power at this position by 5%; if εk is negative, the laser power is reduced by 5% and the pulse frequency is reduced by 10%.

10. A fiber grating production control system for implementing the fiber grating production control method of any one of claims 1-9, characterized by: Comprise: a data acquisition module, which acquires the pre-stress spectrum curve P0 of the target fiber before writing, and a selected set of ultraviolet laser parameters L; a disturbance feature modeling module, which constructs a disturbance sensitive feature vector D0 of the fiber segment according to the pre-stress spectrum curve P0, including multi-scale stress amplitude gradient, frequency distribution entropy and spectrum line phase disturbance index; a disturbance function construction module, which acquires the acoustic emission signal S(t) and laser incident parameter L(t) of the fiber segment in real time during the ultraviolet laser writing process, and calculates the joint disturbance function F(t)=f(S(t),L(t)) in the time window Δt; an event determination module, which dynamically matches the joint disturbance function F(t) with the obtained disturbance sensitive feature vector D0 to determine whether there is a potential nonlinear stress disturbance event Ek, and if there is a potential nonlinear stress disturbance event Ek, it is marked with its corresponding time tk and disturbance intensity Rk; a disturbance prediction adjustment module, which constructs a disturbance prediction model M according to the corresponding time tk and disturbance intensity Rk, to predict the disturbance trend curve R'(t) in the subsequent Δt' time period, and dynamically adjust the laser parameters based on the prediction result; a compensation control module, which acquires the reflection spectrum after the grating writing is completed, and compares it with the target template spectrum; if the correlation coefficient of the two is less than the threshold θ, backtrack the error accumulation path of the disturbance intensity Rk in the event determination module to the disturbance prediction adjustment module, and execute the parameter compensation writing process.

Images