Optical fiber sensing precision enhancement method for high-resolution spectrum subdivision

Abstract

The invention discloses an optical fiber sensing precision enhancement method for high-resolution spectrum subdivision. According to the invention, the spectrum subdivision algorithm based on Fourier transform is applied, so that spectrum subdivision and resolution enhancement are realized, and the optical fiber sensing precision enhancement method based on high-resolution spectrum subdivision has remarkable beneficial effects on the whole optical fiber sensing precision enhancement method based on high-resolution spectrum subdivision. Fourier transform can convert a time domain signal into a frequency domain signal, high-frequency noise and low-frequency interference are effectively removed through frequency domain filtering, and frequency components in a target spectral range are reserved, so that the purity of the spectral signal is improved. And the filtered frequency domain signal is restored into a time domain signal through inverse Fourier transform, so that the integrity of the signal is kept. Furthermore, the time domain signals are subdivided through a linear interpolation algorithm, the data point density is increased, and the spectral resolution is remarkably improved. And a more accurate and detailed data basis is provided for recognition and analysis of subsequent characteristic spectrums, so that the measurement precision and reliability of the optical fiber sensor are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical fiber sensors, and in particular relates to an optical fiber sensing accuracy enhancement method for high-resolution spectral segmentation. Background Art

[0002] With the rapid development of modern industry and science and technology, fiber optic sensors, as an efficient and non-invasive measurement tool, have been widely used in environmental monitoring, medical diagnosis, structural health monitoring and other fields. Fiber optic sensors sense and transmit information about the object being measured through changes in optical signals, and have the advantages of anti-electromagnetic interference, high temperature resistance, and high sensitivity. However, in practical applications, the measurement accuracy and resolution of fiber optic sensors are affected by many factors, such as environmental noise, instrument noise, light source intensity fluctuations, etc. These factors limit the performance and application scope of fiber optic sensors. In order to improve the measurement accuracy and resolution of fiber optic sensors, researchers have proposed a variety of methods, including signal preprocessing, spectral segmentation technology, and characteristic spectrum recognition.

[0003] However, the existing methods still have some shortcomings. Traditional signal preprocessing methods may not be able to completely eliminate noise and interference, resulting in a decrease in signal quality; at the same time, the existing characteristic spectrum recognition and analysis methods may not be accurate enough to establish an accurate quantitative relationship between the spectrum and the measured parameters. Summary of the invention

[0004] The purpose of the present invention is to provide a method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation in order to solve the above-mentioned problem.

[0005] The technical solution adopted by the present invention is as follows: a method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation, the method comprising the following steps:

[0006] S1: Arrange the optical fiber sensor in the area to be tested to ensure that the sensor is in full contact with the object to be tested; use a high-resolution spectrometer to collect the optical signal output by the optical fiber sensor to ensure that the signal is complete and the noise is low;

[0007] S2: De-noising the collected optical signal, using wavelet transform or filter method to eliminate environmental noise and instrument noise; performing signal normalization processing to eliminate the influence of light source intensity fluctuation on the measurement result;

[0008] S3: According to the measured object and measurement requirements, select Fourier transform spectroscopy spectral segmentation technology to achieve high-resolution spectral segmentation;

[0009] S4: Extract spectral data from the preprocessed signal, including key information of intensity and wavelength; ensure the accuracy and consistency of data extraction, and provide a reliable basis for subsequent analysis;

[0010] S5: Apply the selected spectral segmentation technology to process the extracted spectral data to achieve spectral segmentation and resolution enhancement; improve the accuracy and resolution of spectral segmentation through algorithm optimization and parameter adjustment;

[0011] S6: identifying characteristic spectra in the subdivided spectra, performing in-depth analysis on the characteristic spectra, and extracting physical quantities or chemical composition information related to the measured object;

[0012] S7: Based on the characteristic spectrum analysis results, a mathematical model between the output of the optical fiber sensor and the parameters of the object being measured is established; a standard sample or a sample with known parameters is used to calibrate the model to ensure the accuracy and reliability of the model;

[0013] S8: Apply the calibrated model to the real-time monitoring system to achieve high-precision, real-time monitoring of the area to be measured; feed back the monitoring data to the control system or display interface in real time, so that the operator can understand the status of the measured object in a timely manner;

[0014] S9: Continuously optimize the system according to actual application conditions, including sensor layout, signal processing algorithm, and model update; regularly maintain and calibrate the system to ensure long-term stability and measurement accuracy of the system.

[0015] In a preferred embodiment, in step S1, a high-resolution spectrometer is used to collect the optical signal output by the optical fiber sensor, and a suitable sampling rate is set to ensure that the time resolution of the signal meets the requirements; at the same time, the integration time of the spectrometer is adjusted to balance the signal strength and the noise level to ensure that the collected signal is complete, stable and has low noise; and the environmental parameters are recorded so that the influence of environmental factors on the measurement results can be considered in subsequent analysis.

[0016] In a preferred embodiment, in step S2, the signal is first decomposed by wavelet transform method to identify and remove high-frequency noise components and retain low-frequency effective signals; a suitable wavelet basis and decomposition level are selected to achieve effective separation of noise and effective signals; then, a filter is used to further smooth the signal and eliminate the remaining random noise; signal normalization processing is performed to normalize the signal intensity to the same scale to eliminate the influence of light source intensity fluctuations and optical fiber transmission loss factors on the measurement results; during the normalization process, the maximum and minimum values ​​of the signal are calculated and linearly transformed to a specified range to ensure the consistency and comparability of subsequent analysis.

[0017] In a preferred embodiment, in step S3, Fourier transform spectroscopy obtains spectral information through Fourier transform of the interference pattern, which can achieve high-resolution spectral segmentation; the parameters of the Fourier transform, including the number of sampling points of the interference pattern and the maximum optical path difference, are set to balance the resolution and computational complexity; for scenes with rapid dynamic changes or requiring real-time monitoring, wavelength scanning spectroscopy is selected, which obtains spectral information by scanning the wavelength of the monochromator, and sets the scanning speed and step size to meet the requirements of real-time performance and resolution.

[0018] In a preferred embodiment, in step S4, a peak detection algorithm is used to identify the peak position in the spectrum, and the wavelength and intensity value of each peak are recorded; for continuous spectra, an interpolation method is used to smooth the spectral curve to improve data density and accuracy; the accuracy and consistency of data extraction are ensured, and random errors are reduced by taking the average value of multiple measurements; the extracted spectral data should contain sufficient information to reflect the characteristics and changes of the object being measured; the extracted spectral data is stored in a standard format to provide a reliable basis for subsequent analysis and modeling.

[0019] In a preferred embodiment, in step S5, spectrum segmentation and resolution enhancement are performed using a spectrum segmentation algorithm based on Fourier transform, which specifically includes:

[0020] Step S5.1: Fourier transform preprocessing: Perform fast Fourier transform (FFT) on the extracted spectral data to convert the time domain signal into the frequency domain signal. The calculation formula is:

[0021] Among them, F(ν) is the frequency domain signal, f(t) is the time domain signal, ν is the frequency, and t is the time;

[0022] Step S5.2: frequency domain filtering, filtering the signal in the frequency domain to remove high-frequency noise and low-frequency interference; the filter design uses a bandpass filter to retain the frequency components within the target spectral range;

[0023] Step S5.3: Inverse Fourier transform: perform inverse Fourier transform on the filtered frequency domain signal to restore it to the time domain signal. The calculation formula is:

[0024] Among them, f filtered (t) is the restored time domain signal;

[0025] Step S5.4: Spectral subdivision, subdividing the restored time domain signal to improve the spectral resolution; the subdivision method uses a linear interpolation algorithm to increase the data point density, and the calculation formula of linear interpolation is:

[0026]

[0027] Among them, f interpolated (t) is the interpolated signal, t0 and t1 are the positions of the original data points;

[0028] Step S5.5: Parameter optimization, using the gradient boosting decision tree (GBDT) algorithm to optimize the parameters of the Fourier transform and interpolation algorithms to improve the accuracy of spectral segmentation; the optimized parameters include: filter bandwidth, interpolation point density, and window function selection; optimization method: select the optimal parameter combination through experimental verification or simulation analysis;

[0029] Step S5.6: Resolution evaluation: perform resolution evaluation on the subdivided spectrum to ensure that it meets the measurement requirements. The evaluation index is the half-peak width of the spectrum, and the evaluation formula is:

[0030] Where σ is the standard deviation of the Gaussian function, which represents the width of the spectral line;

[0031] Step S5.7: Output the results, and output the segmented high-resolution spectral data to provide data support for subsequent characteristic spectrum identification and analysis.

[0032] In a preferred embodiment, in step S5.5, the specific method includes:

[0033] Step S5.5.1: Data preparation, collecting spectral segmentation data under different parameter combinations, including filter bandwidth, interpolation point density, and window function selection; marking the spectral segmentation effect indicators corresponding to each set of parameters, including resolution and signal-to-noise ratio;

[0034] Step S5.5.2: Feature extraction, extracting features from each set of parameter combinations, including filter characteristics, interpolation point distribution, and window function characteristics

[0035] Step S5.5.3: Model training, using the gradient boosting decision tree (GBDT) algorithm to train the model and establish a mapping relationship between parameter combinations and spectral segmentation effects;

[0036] GBDT learns data through a forest of decision trees, where each tree tries to correct the mistakes of the previous tree.

[0037] The training formula is: GBDT = Train (X, y), where X represents the feature matrix, including filter bandwidth, interpolation point density, and window function characteristics; y represents the target variable, i.e., the spectral segmentation effect index; GBDT represents the trained GBDT model

[0038] Step S5.5.4: Parameter prediction and optimization, using the trained GBDT model to predict the spectral segmentation effect under new parameter combinations;

[0039] The prediction formula is: y^=Predict(Xnew,GBDT), where Xnew represents the new parameter combination feature, and y^ represents the predicted spectral segmentation effect index;

[0040] Step S5.5.5: Gradient boosting and parameter adjustment. According to the prediction results, the gradient boosting method is used to adjust the parameters and find the optimal parameter combination;

[0041] The gradient boosting formula is: Params opt =GradientAscent(y^,Params)

[0042] Params opt Represents the optimal parameter combination

[0043] y^ represents the predicted spectral segmentation effect index;

[0044] Params represents the current parameter combination

[0045] GradientAscent represents the gradient boosting function, which is used to adjust the parameters according to the gradient of the prediction effect indicator.

[0046] Step S5.5.6: Verification and iteration: Use the verification set to verify the optimized parameter combination to ensure its effectiveness and generalization ability; further iterate and optimize the parameters based on the verification results.

[0047] In a preferred embodiment, in step S6, spectral analysis software or a custom algorithm is used to identify the characteristic spectrum according to its shape, position, and intensity characteristics; a partial least squares method is used to establish a quantitative relationship between the characteristic spectrum and the measured parameter to improve the accuracy and reliability of the analysis. The specific algorithm process includes:

[0048] S1. Data preprocessing: center or standardize the spectral data to eliminate the dimension effect; perform the same preprocessing on the measured parameters;

[0049] S2. Extract principal components: extract principal component latent variables from the spectral matrix X and the measured parameter matrix Y at the same time; these latent variables are linear combinations of X and Y, and maximize the covariance of X and Y;

[0050] S3. Establish a regression model: Use the extracted latent variables to establish a regression model between X and Y; determine the number of latent variables through iterative optimization;

[0051] S4. Model validation: Use cross-validation to evaluate the predictive performance of the model; adjust model parameters, including the number of latent variables, to optimize model performance;

[0052] S5. Prediction and analysis: Use the established PLS model to predict new spectral data and obtain estimated values ​​of the measured parameters; analyze and interpret the prediction results;

[0053] For the spectral matrix X (n × p) and the measured parameter matrix Y (n × q), the formula for extracting latent variables t and u is:

[0054] t = Xw;

[0055] u=Yc;

[0056] In the latent variable optimization process, w and c are iteratively optimized to maximize the covariance of t and u: maxCov(t,u)=maxCov(Xw,Yc);

[0057] The formula for establishing a regression model between t and u is:

[0058] u=bt+e;

[0059] For the new spectral data x_new, the corresponding predicted value formula of the measured parameter y_new is: y_new=x_newW(B^T)Q^T+f;

[0060] In the above formula:

[0061] X represents the spectrum matrix, with n rows and p columns, representing the spectrum intensity of n samples at p wavelengths;

[0062] Y represents the measured parameter matrix, with n rows and q columns, representing the q measured parameters of n samples;

[0063] t represents the latent variable of X, indicating the main change direction of the spectral data;

[0064] u represents the latent variable of Y, indicating the main direction of change of the measured parameter;

[0065] w represents the weight vector of X, used to extract t;

[0066] c represents the weight vector of Y, which is used to extract u;

[0067] b represents the regression coefficient, which indicates the linear relationship between t and u;

[0068] e represents the error term, which represents the variation that the model cannot explain;

[0069] W represents the weight matrix of X, which contains the weight vectors of all latent variables;

[0070] B represents the regression coefficient matrix, which contains the regression coefficients corresponding to all latent variables;

[0071] Q represents the weight matrix of Y, which contains the weight vectors corresponding to all latent variables;

[0072] f represents the constant term and represents the intercept of the model.

[0073] In a preferred embodiment, in step S7, standard samples or samples with known parameters are used for model training, and the model parameters are adjusted by an optimization algorithm to minimize the prediction error; the model is calibrated, and the model performance is evaluated using an independent validation set, and the evaluation index is calculated to ensure the accuracy and reliability of the model.

[0074] In a preferred embodiment, in step S9, the monitoring data is transmitted to the control system or display interface in real time, and data visualization technology is used to intuitively display the status and change trend of the object under test; an alarm threshold is set, and when the monitoring data exceeds the normal range, an alarm signal is issued in time to notify the operator to take corresponding measures.

[0075] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0076] 1. In the present invention, by applying a spectrum subdivision algorithm based on Fourier transform, spectrum subdivision and resolution enhancement are achieved, which has a significant beneficial effect on the fiber optic sensing accuracy enhancement method of the entire high-resolution spectrum subdivision. First, Fourier transform can convert time domain signals into frequency domain signals, effectively remove high-frequency noise and low-frequency interference through frequency domain filtering, retain the frequency components within the target spectrum range, and thus improve the purity of the spectrum signal. Then, the inverse Fourier transform restores the filtered frequency domain signal to the time domain signal, maintaining the integrity of the signal. Further, the time domain signal is subdivided by a linear interpolation algorithm, which increases the data point density and significantly improves the spectral resolution. In addition, the parameters of the Fourier transform and interpolation algorithms are optimized using the GBDT algorithm to ensure that the accuracy and resolution of the spectrum subdivision are optimal. These processes not only improve the quality of the spectral data, but also provide a more accurate and detailed data basis for the subsequent identification and analysis of characteristic spectra, thereby enhancing the measurement accuracy and reliability of the fiber optic sensor.

[0077] 2. In the present invention, the partial least squares method (PLS) is used to establish the quantitative relationship between the characteristic spectrum and the measured parameter. This step also has a significant beneficial effect on the fiber optic sensing accuracy enhancement method of the entire high-resolution spectrum segmentation. Through spectral analysis software or custom algorithms, accurate identification is performed according to the characteristics such as the shape, position, and intensity of the characteristic spectrum, and physical quantities or chemical composition information related to the measured object can be effectively extracted. The PLS method simultaneously extracts the principal components of the spectral matrix and the measured parameter matrix, establishes a regression model between the two, maximizes the covariance between the spectral data and the measured parameters, and thus improves the prediction performance and interpretation ability of the model. Model verification and parameter optimization ensure the accuracy and reliability of the model, so that the analysis results based on the characteristic spectrum can more accurately reflect the state and changes of the measured object. This step not only improves the data analysis capability of the fiber optic sensor, but also provides strong support for its application in real-time monitoring and control systems, and achieves the goal of high-precision and real-time monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a schematic diagram of the process principle of the present invention. DETAILED DESCRIPTION

[0079] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0080] Reference Figure 1 ,

[0081] A method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation, the method comprising the following steps:

[0082] S1: Arrange the fiber optic sensor in the area to be tested, and ensure that the sensor is in full contact with the object to be tested. Use a high-resolution spectrometer to collect the optical signal output by the fiber optic sensor to ensure that the signal is complete and has low noise;

[0083] S2: De-noising the collected optical signal, using wavelet transform or filter to eliminate environmental noise and instrument noise; performing signal normalization to eliminate the influence of light source intensity fluctuation on the measurement result;

[0084] S3: According to the measured object and measurement requirements, select Fourier transform spectroscopy spectral segmentation technology to achieve high-resolution spectral segmentation;

[0085] S4: Extract spectral data from the preprocessed signal, including key information such as intensity and wavelength. Ensure the accuracy and consistency of data extraction to provide a reliable basis for subsequent analysis;

[0086] S5: Apply the selected spectral segmentation technology to process the extracted spectral data to achieve spectral segmentation and resolution enhancement. Improve the accuracy and resolution of spectral segmentation through algorithm optimization and parameter adjustment;

[0087] S6: Identify characteristic spectra, such as absorption peaks, emission peaks, etc. in the segmented spectrum. Perform in-depth analysis on the characteristic spectrum to extract physical quantities or chemical composition information related to the object being measured;

[0088] S7: Based on the characteristic spectrum analysis results, a mathematical model between the output of the optical fiber sensor and the parameters of the object being measured is established. The model is calibrated using standard samples or samples with known parameters to ensure the accuracy and reliability of the model;

[0089] S8: Apply the calibrated model to the real-time monitoring system to achieve high-precision, real-time monitoring of the area to be measured. Feedback the monitoring data to the control system or display interface in real time, so that the operator can understand the status of the measured object in a timely manner;

[0090] S9: Continuously optimize the system according to actual application conditions, including sensor layout, signal processing algorithm, model update, etc. Regularly maintain and calibrate the system to ensure long-term stability and measurement accuracy.

[0091] In step S1, a high-resolution spectrometer (e.g., a spectrometer with a spectral resolution of nanometers) is used to collect the optical signal output by the optical fiber sensor, and a suitable sampling rate is set to ensure that the time resolution of the signal meets the requirements. At the same time, the integration time of the spectrometer is adjusted to balance the signal intensity and the noise level to ensure that the collected signal is complete, stable and has low noise. Environmental parameters such as temperature and humidity are recorded so that the influence of environmental factors on the measurement results can be considered in subsequent analysis.

[0092] In the step S2, the wavelet transform method is first used to decompose the signal, identify and remove high-frequency noise components, and retain low-frequency effective signals. Select a suitable wavelet basis and decomposition level to achieve effective separation of noise and effective signals. Then, use a filter (such as a low-pass filter) to further smooth the signal and eliminate the remaining random noise. Perform signal normalization processing, normalize the signal intensity to the same scale, and eliminate the influence of factors such as light source intensity fluctuations and optical fiber transmission losses on the measurement results. During the normalization process, the maximum and minimum values ​​of the signal are calculated, and it is linearly transformed to a specified range (such as 0 to 1) to ensure the consistency and comparability of subsequent analysis.

[0093] In step S3, Fourier transform spectroscopy obtains spectral information through Fourier transform of interference pattern, and can achieve high-resolution spectral segmentation. Set the parameters of Fourier transform, such as the number of sampling points of interference pattern, the maximum optical path difference, etc., to balance the resolution and computational complexity. For scenes with fast dynamic changes or real-time monitoring, wavelength scanning spectroscopy is selected, which obtains spectral information by scanning the wavelength of the monochromator, and sets the scanning speed and step size to meet the requirements of real-time performance and resolution.

[0094] In step S4, a peak detection algorithm is used to identify the peak position in the spectrum, and the wavelength and intensity value of each peak are recorded. For continuous spectra, an interpolation method is used to smooth the spectral curve to improve data density and accuracy. To ensure the accuracy and consistency of data extraction, random errors are reduced by taking the average value through multiple measurements. The extracted spectral data should contain enough information to reflect the characteristics and changes of the object being measured. The extracted spectral data is stored in a standard format to provide a reliable basis for subsequent analysis and modeling.

[0095] In step S5, spectrum segmentation and resolution enhancement are performed using a spectrum segmentation algorithm based on Fourier transform, which specifically includes:

[0096] Step S5.1: Fourier transform preprocessing: Perform fast Fourier transform (FFT) on the extracted spectral data to convert the time domain signal into the frequency domain signal. The calculation formula is:

[0097] Among them, F(ν) is the frequency domain signal, f(t) is the time domain signal, ν is the frequency, and t is the time.

[0098] Step S5.2: Frequency domain filtering, filtering the signal in the frequency domain to remove high-frequency noise and low-frequency interference. The filter design uses a bandpass filter to retain the frequency components within the target spectral range;

[0099] Step S5.3: Inverse Fourier transform, perform inverse Fourier transform (IFFT) on the filtered frequency domain signal to restore it to the time domain signal. The calculation formula is:

[0100] Among them, f filtered (t) is the restored time domain signal;

[0101] Step S5.4: Spectral subdivision: subdividing the restored time domain signal to improve the spectral resolution. The subdivision method uses a linear interpolation algorithm to increase the data point density. The calculation formula of linear interpolation is:

[0102]

[0103] Among them, f interpolated(t) is the interpolated signal, and t0 and t1 are the positions of the original data points.

[0104] Step S5.5: Parameter optimization, using the gradient boosting decision tree (GBDT) algorithm to optimize the parameters of the Fourier transform and interpolation algorithms to improve the accuracy of spectral segmentation. The optimization parameters include: filter bandwidth, interpolation point density, and window function selection. Optimization method: select the optimal parameter combination through experimental verification or simulation analysis;

[0105] Step S5.6: Resolution evaluation: perform resolution evaluation on the subdivided spectrum to ensure that the measurement requirements are met. The evaluation index is the spectrum half-maximum width (FWHM), where the evaluation formula is:

[0106] Where σ is the standard deviation of the Gaussian function, which represents the width of the spectral line;

[0107] Step S5.7: Output the results, output the segmented high-resolution spectral data, and provide data support for subsequent characteristic spectrum identification and analysis

[0108] In the step S5.5, the specific method includes:

[0109] Step S5.5.1: Data preparation, collect spectral segmentation data under different parameter combinations, including filter bandwidth, interpolation point density, window function selection, etc. Mark the spectral segmentation effect indicators corresponding to each set of parameters, such as resolution, signal-to-noise ratio, etc.

[0110] Step S5.5.2: Feature extraction, extracting features from each set of parameter combinations, including filter characteristics, interpolation point distribution, and window function characteristics

[0111] Step S5.5.3: Model training, using the gradient boosted decision tree (GBDT) algorithm to train the model and establish a mapping relationship between parameter combinations and spectral segmentation effects.

[0112] GBDT learns data through a forest of decision trees, where each tree tries to correct the mistakes of the previous tree.

[0113] The training formula is: GBDT = Train (X, y), where X represents the feature matrix, including filter bandwidth, interpolation point density, window function characteristics and other features; y represents the target variable, i.e., the spectral segmentation effect index; GBDT represents the trained GBDT model

[0114] Step S5.5.4: Parameter prediction and optimization, using the trained GBDT model to predict the spectral segmentation effect under new parameter combinations.

[0115] The prediction formula is: y^=Predict(Xnew,GBDT), where Xnew represents the new parameter combination feature, and y^ represents the predicted spectral segmentation effect index;

[0116] Step S5.5.5: Gradient boosting and parameter adjustment. According to the prediction results, the gradient boosting method is used to adjust the parameters and find the optimal parameter combination.

[0117] The gradient boosting formula is: Params opt =GradientAscent(y^,Params)

[0118] Params opt Represents the optimal parameter combination

[0119] y^ represents the predicted spectral segmentation effect index;

[0120] Params represents the current parameter combination

[0121] GradientAscent represents the gradient boosting function, which is used to adjust the parameters according to the gradient of the prediction effect indicator.

[0122] Step S5.5.6: Verification and iteration: Use the validation set to verify the optimized parameter combination to ensure its effectiveness and generalization ability. Further iterate and optimize the parameters based on the validation results.

[0123] Through the above steps, the GBDT algorithm is used to realize the adaptive optimization of Fourier transform and interpolation algorithm parameters, thereby improving the accuracy and resolution of spectral segmentation.

[0124] In step S6, spectral analysis software or a custom algorithm is used to identify the characteristic spectrum based on its shape, position, intensity and other characteristics. The partial least squares method is used to establish a quantitative relationship between the characteristic spectrum and the measured parameter to improve the accuracy and reliability of the analysis. The specific algorithm process includes:

[0125] S1. Data preprocessing: Center or standardize the spectral data to eliminate the dimension effect. Perform the same preprocessing on the measured parameters.

[0126] S2. Extract principal components: Extract principal components, i.e. latent variables (LVs), from both the spectral matrix X and the measured parameter matrix Y. These latent variables are linear combinations of X and Y and maximize the covariance of X and Y.

[0127] S3. Establish a regression model: Use the extracted latent variables to establish a regression model between X and Y. The number of latent variables is determined through iterative optimization.

[0128] S4. Model validation: Use methods such as cross-validation to evaluate the predictive performance of the model. Adjust model parameters, such as the number of latent variables, to optimize model performance.

[0129] S5. Prediction and analysis: Use the established PLS model to predict new spectral data and obtain the estimated values ​​of the measured parameters. Analyze and interpret the prediction results;

[0130] For the spectral matrix X (n × p) and the measured parameter matrix Y (n × q), the formula for extracting latent variables t and u is:

[0131] t = Xw;

[0132] u=Yc;

[0133] In the latent variable optimization process, w and c are iteratively optimized to maximize the covariance of t and u: maxCov(t,u)=maxCov(Xw,Yc);

[0134] The formula for establishing a regression model between t and u is:

[0135] u=bt+e;

[0136] For the new spectral data x_new, the corresponding predicted value formula of the measured parameter y_new is: y_new=x_newW(B^T)Q^T+f;

[0137] In the above formula:

[0138] X represents the spectrum matrix, with n rows and p columns, representing the spectrum intensity of n samples at p wavelengths;

[0139] Y represents the measured parameter matrix, with n rows and q columns, representing the q measured parameters of n samples;

[0140] t represents the latent variable of X, indicating the main change direction of the spectral data;

[0141] u represents the latent variable of Y, indicating the main direction of change of the measured parameter;

[0142] w represents the weight vector of X, used to extract t;

[0143] c represents the weight vector of Y, which is used to extract u;

[0144] b represents the regression coefficient, which indicates the linear relationship between t and u;

[0145] e represents the error term, which represents the variation that the model cannot explain;

[0146] W represents the weight matrix of X, which contains the weight vectors of all latent variables;

[0147] B represents the regression coefficient matrix, which contains the regression coefficients corresponding to all latent variables;

[0148] Q represents the weight matrix of Y, which contains the weight vectors corresponding to all latent variables;

[0149] f represents the constant term and represents the intercept of the model.

[0150] In step S7, a standard sample or a sample with known parameters is used for model training, and the model parameters are adjusted by an optimization algorithm (such as gradient descent method) to minimize the prediction error. The model is calibrated, and the model performance is evaluated using an independent validation set, and evaluation indicators (such as mean square error, determination coefficient, etc.) are calculated to ensure the accuracy and reliability of the model. The model is iteratively optimized according to the calibration results until the measurement requirements are met.

[0151] In step S9, the monitoring data is transmitted to the control system or display interface in real time, and data visualization technology (such as real-time curves, dashboards, etc.) is used to intuitively display the state and change trend of the object under test. An alarm threshold is set, and when the monitoring data exceeds the normal range, an alarm signal is issued in time to notify the operator to take corresponding measures. A data storage and query system is established to record historical monitoring data for subsequent analysis and tracing.

[0152] In the present invention, by applying a spectrum subdivision algorithm based on Fourier transform, spectrum subdivision and resolution enhancement are achieved, which has a significant beneficial effect on the entire high-resolution spectrum subdivision fiber optic sensing accuracy enhancement method. First, Fourier transform can convert time domain signals into frequency domain signals, effectively remove high-frequency noise and low-frequency interference through frequency domain filtering, retain the frequency components within the target spectrum range, and thus improve the purity of the spectrum signal. Then, the inverse Fourier transform restores the filtered frequency domain signal to the time domain signal, maintaining the integrity of the signal. Further, the time domain signal is subdivided by a linear interpolation algorithm, which increases the data point density and significantly improves the spectral resolution. In addition, the parameters of the Fourier transform and interpolation algorithms are optimized using the GBDT algorithm to ensure that the accuracy and resolution of the spectrum subdivision are optimal. These processes not only improve the quality of the spectral data, but also provide a more accurate and detailed data basis for the subsequent identification and analysis of characteristic spectra, thereby enhancing the measurement accuracy and reliability of the fiber optic sensor.

[0153] In the present invention, the partial least squares method (PLS) is used to establish the quantitative relationship between the characteristic spectrum and the measured parameter. This step also has a significant beneficial effect on the fiber optic sensing accuracy enhancement method of the entire high-resolution spectrum segmentation. Through spectral analysis software or a custom algorithm, accurate identification is performed according to the characteristics such as the shape, position, and intensity of the characteristic spectrum, and physical quantities or chemical composition information related to the measured object can be effectively extracted. The PLS method simultaneously extracts the principal components of the spectral matrix and the measured parameter matrix, establishes a regression model between the two, maximizes the covariance between the spectral data and the measured parameters, and thus improves the prediction performance and interpretation ability of the model. Model verification and parameter optimization ensure the accuracy and reliability of the model, so that the analysis results based on the characteristic spectrum can more accurately reflect the state and changes of the measured object. This step not only improves the data analysis capability of the fiber optic sensor, but also provides strong support for its application in real-time monitoring and control systems, and achieves the goal of high-precision and real-time monitoring.

[0154] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0155] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation, characterized in that: The method comprises the following steps: S1: Arrange the optical fiber sensor in the area to be tested to ensure that the sensor is in full contact with the object to be tested; use a high-resolution spectrometer to collect the optical signal output by the optical fiber sensor to ensure that the signal is complete and the noise is low; S2: De-noising the collected optical signal, using wavelet transform or filter method to eliminate environmental noise and instrument noise; performing signal normalization processing to eliminate the influence of light source intensity fluctuation on the measurement result; S3: According to the measured object and measurement requirements, select Fourier transform spectroscopy spectral segmentation technology to achieve high-resolution spectral segmentation; S4: Extract spectral data from the preprocessed signal, including key information of intensity and wavelength; ensure the accuracy and consistency of data extraction, and provide a reliable basis for subsequent analysis; S5: Apply the selected spectral segmentation technology to process the extracted spectral data to achieve spectral segmentation and resolution enhancement; improve the accuracy and resolution of spectral segmentation through algorithm optimization and parameter adjustment; S6: identifying characteristic spectra in the subdivided spectra, performing in-depth analysis on the characteristic spectra, and extracting physical quantities or chemical composition information related to the measured object; S7: Based on the characteristic spectrum analysis results, a mathematical model between the output of the optical fiber sensor and the parameters of the object being measured is established; a standard sample or a sample with known parameters is used to calibrate the model to ensure the accuracy and reliability of the model; S8: Apply the calibrated model to the real-time monitoring system to achieve high-precision, real-time monitoring of the area to be measured; feed back the monitoring data to the control system or display interface in real time, so that the operator can understand the status of the measured object in a timely manner; S9: Continuously optimize the system according to actual application conditions, including sensor layout, signal processing algorithm, and model update; regularly maintain and calibrate the system to ensure long-term stability and measurement accuracy of the system.

2. The method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation according to claim 1, characterized in that: In step S1, a high-resolution spectrometer is used to collect the optical signal output by the optical fiber sensor, and a suitable sampling rate is set to ensure that the time resolution of the signal meets the requirements; at the same time, the integration time of the spectrometer is adjusted to balance the signal strength and the noise level, so as to ensure that the collected signal is complete, stable and has low noise; Record environmental parameters so that the impact of environmental factors on measurement results can be considered in subsequent analysis.

3. The method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation according to claim 1, characterized in that: In the step S2, the signal is first decomposed by wavelet transform method, high-frequency noise components are identified and removed, and low-frequency effective signals are retained; a suitable wavelet basis and decomposition level are selected to achieve effective separation of noise and effective signals; then, a filter is used to further smooth the signal and eliminate the remaining random noise; signal normalization processing is performed to normalize the signal intensity to the same scale to eliminate the influence of light source intensity fluctuation and optical fiber transmission loss factors on the measurement results; During the normalization process, the maximum and minimum values ​​of the signal are calculated and linearly transformed to the specified range to ensure the consistency and comparability of subsequent analysis.

4. The method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation according to claim 1, characterized in that: In step S3, Fourier transform spectroscopy obtains spectral information through Fourier transform of the interference pattern to achieve high-resolution spectral segmentation; Fourier transform parameters are set, including the number of sampling points of the interference pattern and the maximum optical path difference, to balance resolution and computational complexity; for scenes with rapid dynamic changes or requiring real-time monitoring, wavelength scanning spectroscopy is selected, which obtains spectral information by scanning the wavelength of the monochromator, and sets the scanning speed and step size to meet the requirements of real-time performance and resolution.

5. The method for enhancing the accuracy of optical fiber sensing with high resolution spectrum segmentation according to claim 1, characterized in that: In step S4, a peak detection algorithm is used to identify the peak position in the spectrum, and the wavelength and intensity value of each peak are recorded; for continuous spectra, an interpolation method is used to smooth the spectrum curve to improve data density and accuracy; the accuracy and consistency of data extraction are ensured, and random errors are reduced by taking the average value through multiple measurements; the extracted spectral data should contain sufficient information to reflect the characteristics and changes of the object being measured; the extracted spectral data is stored in a standard format to provide a reliable basis for subsequent analysis and modeling.

6. The method for enhancing the accuracy of optical fiber sensing with high-resolution spectral segmentation according to claim 1, characterized in that: In step S5, spectrum segmentation and resolution enhancement are performed using a spectrum segmentation algorithm based on Fourier transform, which specifically includes: Step S5.1: Fourier transform preprocessing: Perform fast Fourier transform (FFT) on the extracted spectral data to convert the time domain signal into the frequency domain signal. The calculation formula is: Among them, F(ν) is the frequency domain signal, f(t) is the time domain signal, ν is the frequency, and t is the time; Step S5.2: frequency domain filtering, filtering the signal in the frequency domain to remove high-frequency noise and low-frequency interference; the filter design uses a bandpass filter to retain the frequency components within the target spectral range; Step S5.3: Inverse Fourier transform: perform inverse Fourier transform on the filtered frequency domain signal to restore it to the time domain signal. The calculation formula is: Among them, f filtered (t) is the restored time domain signal; Step S5.4: Spectral subdivision, subdividing the restored time domain signal to improve the spectral resolution; the subdivision method uses a linear interpolation algorithm to increase the data point density, and the calculation formula of linear interpolation is: Among them, f interpolated (t) is the interpolated signal, t0 and t1 are the positions of the original data points; Step S5.5: Parameter optimization, using the gradient boosting decision tree (GBDT) algorithm to optimize the parameters of the Fourier transform and interpolation algorithms to improve the accuracy of spectral segmentation; the optimized parameters include: filter bandwidth, interpolation point density, and window function selection; optimization method: select the optimal parameter combination through experimental verification or simulation analysis; Step S5.6: Resolution evaluation: perform resolution evaluation on the subdivided spectrum to ensure that it meets the measurement requirements. The evaluation index is the half-peak width of the spectrum, and the evaluation formula is: Where σ is the standard deviation of the Gaussian function, which represents the width of the spectral line; Step S5.7: Output the results, and output the segmented high-resolution spectral data to provide data support for subsequent characteristic spectrum identification and analysis.

7. The method for enhancing the accuracy of optical fiber sensing with high resolution spectrum segmentation according to claim 6, characterized in that: In the step S5.5, the specific method includes: Step S5.5.1: Data preparation, collecting spectral segmentation data under different parameter combinations, including filter bandwidth, interpolation point density, and window function selection; marking the spectral segmentation effect indicators corresponding to each set of parameters, including resolution and signal-to-noise ratio; Step S5.5.2: Feature extraction, extracting features from each set of parameter combinations, including filter characteristics, interpolation point distribution, and window function characteristics Step S5.5.3: Model training, using the gradient boosting decision tree (GBDT) algorithm to train the model and establish a mapping relationship between parameter combinations and spectral segmentation effects; GBDT learns data through a forest of decision trees, where each tree tries to correct the mistakes of the previous tree. The training formula is: GBDT = Train (X, y), where X represents the feature matrix, including filter bandwidth, interpolation point density, and window function characteristics; y represents the target variable, i.e., the spectral segmentation effect index; GBDT represents the trained GBDT model Step S5.5.4: Parameter prediction and optimization, using the trained GBDT model to predict the spectral segmentation effect under new parameter combinations; The prediction formula is: y^=Predict(Xnew,GBDT), where Xnew represents the new parameter combination feature, and y^ represents the predicted spectral segmentation effect index; Step S5.5.5: Gradient boosting and parameter adjustment. According to the prediction results, the gradient boosting method is used to adjust the parameters and find the optimal parameter combination; The gradient boosting formula is: Params opt =GradientAscent(y^,Params) Params opt Represents the optimal parameter combination y^ represents the predicted spectral segmentation effect index; Params represents the current parameter combination GradientAscent represents the gradient boosting function, which is used to adjust the parameters according to the gradient of the prediction effect indicator. Step S5.5.6: Verification and iteration: Use the verification set to verify the optimized parameter combination to ensure its effectiveness and generalization ability; further iterate and optimize the parameters based on the verification results.

8. The method for enhancing the accuracy of optical fiber sensing with high resolution spectrum segmentation according to claim 1, characterized in that: In step S6, spectrum analysis software or a custom algorithm is used to identify the characteristic spectrum according to its shape, position, and intensity characteristics; partial least squares method is used to establish a quantitative relationship between the characteristic spectrum and the measured parameter to improve the accuracy and reliability of the analysis. The specific algorithm process includes: S1. Data preprocessing: center or standardize the spectral data to eliminate the dimension effect; perform the same preprocessing on the measured parameters; S2. Extract principal components: extract principal component latent variables from the spectral matrix X and the measured parameter matrix Y at the same time; these latent variables are linear combinations of X and Y, and maximize the covariance of X and Y; S3. Establish a regression model: Use the extracted latent variables to establish a regression model between X and Y; determine the number of latent variables through iterative optimization; S4. Model validation: Use cross-validation to evaluate the predictive performance of the model; adjust model parameters, including the number of latent variables, to optimize model performance; S5. Prediction and analysis: Use the established PLS model to predict new spectral data and obtain estimated values ​​of the measured parameters; analyze and interpret the prediction results; For the spectral matrix X (n × p) and the measured parameter matrix Y (n × q), the formula for extracting latent variables t and u is: t = Xw; u=Yc; In the latent variable optimization process, w and c are iteratively optimized to maximize the covariance of t and u: max Cov(t,u)=max Cov(Xw,Yc); The formula for establishing a regression model between t and u is: u=bt+e; For the new spectral data x_new, the corresponding predicted value formula of the measured parameter y_new is: y_new=x_newW(B^T)Q^T+f; In the above formula: X represents the spectrum matrix, with n rows and p columns, representing the spectrum intensity of n samples at p wavelengths; Y represents the measured parameter matrix, with n rows and q columns, representing the q measured parameters of n samples; t represents the latent variable of X, indicating the main change direction of the spectral data; u represents the latent variable of Y, indicating the main direction of change of the measured parameter; w represents the weight vector of X, used to extract t; c represents the weight vector of Y, which is used to extract u; b represents the regression coefficient, which indicates the linear relationship between t and u; e represents the error term, which represents the variation that the model cannot explain; W represents the weight matrix of X, which contains the weight vectors of all latent variables; B represents the regression coefficient matrix, which contains the regression coefficients corresponding to all latent variables; Q represents the weight matrix of Y, which contains the weight vectors corresponding to all latent variables; f represents the constant term and represents the intercept of the model.

9. The method for enhancing the accuracy of optical fiber sensing with high resolution spectrum segmentation according to claim 1, characterized in that: In step S7, standard samples or samples with known parameters are used for model training, and model parameters are adjusted through optimization algorithms to minimize prediction errors; model calibration is performed, and model performance is evaluated using an independent validation set, and evaluation indicators are calculated to ensure the accuracy and reliability of the model.

10. The method for enhancing the accuracy of optical fiber sensing with high resolution spectrum segmentation according to claim 1, characterized in that: In step S9, the monitoring data is transmitted to the control system or display interface in real time, and the state and change trend of the measured object are intuitively displayed using data visualization technology; an alarm threshold is set, and when the monitoring data exceeds the normal range, an alarm signal is issued in time to notify the operator to take corresponding measures.

Images