Optical fiber temperature measurement method and system

Abstract

The present invention relates to the field of optical fiber sensing technology, and in particular to an optical fiber temperature measurement method and system. The method comprises: utilizing a multi-scale dynamically modulated and coded laser signal to inject the optical fiber to be measured and the reference optical fiber respectively, collecting the two optical signals and then generating frequency domain data containing the modulation coding characteristics through fast Fourier transform; constructing a virtual reference channel, fusing pre-stored historical data through digital matching filtering and self-organizing mapping to achieve stable output of the reference signal; utilizing an extended Kalman filter combined with fuzzy logic to adaptively calibrate the measured data, and outputting calibrated phase data that accurately reflects temperature changes; and finally, outputting high-precision temperature distribution data based on a temperature inversion model and multidimensional correlation analysis. The present invention significantly improves temperature measurement accuracy and system robustness, and is suitable for fields such as industrial and environmental monitoring.

Description

Technical Field

[0001] The present invention relates to the field of optical fiber sensing technology, and in particular to an optical fiber temperature measurement method and system. Background Art

[0002] Fiber-optic temperature measurement technology leverages the inherent properties of optical fiber, such as high-temperature resistance, corrosion resistance, and electromagnetic interference resistance, and plays a vital role in fields such as power generation, petroleum, civil engineering, and disaster monitoring. Traditional fiber-optic temperature measurement methods primarily rely on inputting laser pulses into the fiber, collecting the anti-Stokes (AST) and Stokes (ST) signals, and using the ratio of their light intensities to reflect temperature changes.

[0003] However, existing technologies generally require that the winding length of the optical fiber at the temperature measurement point reach a minimum physical distance to ensure temperature measurement accuracy. When the object to be measured is small or the environment requires a compact design in actual applications, insufficient optical fiber winding length often causes the ratio of AST to ST signals to change nonlinearly due to transmission attenuation, thereby introducing measurement errors and limiting the effective detection distance of the entire optical fiber.

[0004] Furthermore, traditional methods typically employ simple signal ratio compensation techniques or rely on deep learning models to compensate for collected data. However, these methods suffer from deficiencies such as insufficient training data, poor model generalization, and sensitivity to environmental noise, making it difficult to meet high temperature measurement accuracy requirements in practical applications. Given this background, there is an urgent need for a new fiber-optic temperature measurement technology that can overcome the traditional methods' stringent requirements for fiber winding length while enabling real-time, accurate detection of temperature changes in complex environments. This extends the effective detection range of the fiber and improves the overall robustness and stability of the system. Summary of the Invention

[0005] To address the numerous issues with the aforementioned existing technologies, the present invention provides a fiber-optic temperature measurement method and system. This method embeds multi-scale modulation coding features within the laser signal, constructs a virtual reference channel using reference fiber data, and then implements adaptive calibration using an extended Kalman filter combined with fuzzy logic. Ultimately, the calibrated phase data is converted into a temperature value using a temperature inversion model. This method effectively eliminates the signal attenuation problem associated with insufficient fiber winding length in traditional fiber-optic temperature measurement. Furthermore, closed-loop feedback control enables dynamic adaptive adjustment of system parameters, thereby obtaining high-precision temperature distribution data.

[0006] An optical fiber temperature measurement method comprises the following steps:

[0007] The laser signal with multi-scale dynamic modulation coding is injected into the optical fiber to be tested and the reference optical fiber respectively, and the optical signal of each channel is collected and then subjected to fast Fourier transform to generate frequency domain data containing modulation coding characteristics;

[0008] A virtual reference channel is constructed, wherein the reference fiber frequency domain data is demodulated and denoised, and the demodulated data is fused with pre-stored historical data to form virtual reference frequency domain data. The frequency domain data of the fiber under test are then compared with the virtual reference frequency domain data at the same sampling position, and adaptive calibration is performed using an adaptive calibration algorithm. The adaptive calibration algorithm uses an extended Kalman filter combined with fuzzy logic to correct the fiber under test signal to obtain calibrated phase data reflecting temperature changes.

[0009] The temperature inversion of the calibrated phase data is performed based on the pre-established temperature inversion model, and the preliminary temperature data is optimized using multidimensional correlation analysis to output high-precision temperature distribution data.

[0010] Preferably, in the multi-scale dynamic modulation coding step, a high-speed digital signal processor is used to perform sinusoidal wave modulation on the laser output signal, and the frequency of the sinusoidal wave modulation is set to not less than 1000 Hz and not more than 5000 Hz. At the same time, a pulse signal is superimposed by a dedicated pulse coding circuit, and the pulse width is set to not less than 5 microseconds and not more than 50 microseconds. The pulse repetition period is determined according to the temperature response characteristics of the environment to be measured, thereby generating a laser signal containing a fixed frequency component and a pulse coding component in the spectrum, and the laser signal is used to extract the temperature change mark in the subsequent frequency domain data.

[0011] Preferably, in the step of performing fast Fourier transform after collecting the optical signal, Hanning window weighting processing is first applied to the collected time domain data, and then Fourier transform is performed using a digital signal processing algorithm to generate frequency domain data, thereby ensuring that the modulation coding characteristics are retained in the frequency domain data.

[0012] Preferably, in the step of constructing a virtual reference channel, the frequency domain data collected by the reference optical fiber is demodulated and noise suppressed using a digital matched filter, and the demodulated data is fused with the pre-stored historical reference data using a self-organizing map algorithm combined with a fuzzy clustering algorithm to form virtual reference frequency domain data reflecting the ideal reference state.

[0013] Preferably, in the adaptive calibration step, the frequency domain data of the optical fiber to be measured is compared with the virtual reference frequency domain data at the same sampling position to calculate the initial phase error signal, and the extended Kalman filter combined with fuzzy logic is used to recursively predict and update the initial phase error signal, and at the same time, the process noise covariance and the measurement noise covariance are adaptively adjusted online to obtain calibrated phase data that accurately reflects temperature changes.

[0014] Preferably, the temperature inversion step adopts a temperature inversion model established based on offline calibration data, and the temperature inversion model adopts quadratic polynomial fitting to convert the calibrated phase data into temperature values, thereby generating preliminary temperature data.

[0015] Preferably, in the multidimensional correlation analysis step, cross-spectrum analysis is performed on the preliminary temperature data and the modulation coding feature data, the nonlinear correlation between the two is extracted through normalization processing and correlation coefficient calculation, and the preliminary temperature data is robustly optimized using a random sampling consistency algorithm to obtain the corrected temperature data.

[0016] Preferably, the calibrated phase data is monitored in real time through a closed-loop feedback system, which transmits the filtering residuals generated during the calibration process to the multi-scale dynamic modulation coding and virtual reference channel construction steps, and uses an online minimum mean square error algorithm to perform dynamic adaptive adjustment of the modulation parameters and virtual reference construction parameters.

[0017] Preferably, the temperature inversion step further includes iteratively adjusting the temperature inversion model parameters using an online minimum mean square error algorithm, wherein the online minimum mean square error algorithm uses the error between the temperature value obtained by converting the auxiliary temperature reference data and the calibrated phase data as feedback to reduce the error and improve the accuracy of the temperature inversion.

[0018] An optical fiber temperature measurement system for implementing the optical fiber temperature measurement method, the system comprising:

[0019] A laser module, configured to output a laser signal that has undergone multi-scale dynamic modulation encoding;

[0020] An optical path distribution module, configured to equally divide the modulated laser signal and inject it into the optical fiber to be tested and the reference optical fiber respectively. The optical fiber to be tested is used to transmit the signal to the object to be tested, and the reference optical fiber is placed in a temperature-controlled and low-vibration environment.

[0021] A signal detection module is used to collect the return optical signals of the optical fiber to be tested and the reference optical fiber, and convert the optical signals into digital time domain data;

[0022] A digital signal processing module, configured to perform a fast Fourier transform on the digital time domain data, and generate frequency domain data containing modulation coding features after Hanning window weighting processing;

[0023] A virtual reference channel construction module is used to demodulate and suppress noise on the frequency domain data collected by the reference optical fiber, and fuse the demodulated data with pre-stored historical reference data to form virtual reference frequency domain data reflecting an ideal reference state;

[0024] The adaptive calibration module is used to compare the frequency domain data of the optical fiber to be tested with the virtual reference frequency domain data at the same sampling position, recursively predict and update the initial phase error signal using an extended Kalman filter combined with fuzzy logic, and perform online adaptive adjustment of the process noise covariance and the measurement noise covariance to output calibrated phase data reflecting temperature changes;

[0025] a temperature inversion module, configured to convert the calibrated phase data into temperature values ​​based on a pre-established temperature inversion model, and optimize the preliminary temperature data using multidimensional correlation analysis to output high-precision temperature distribution data;

[0026] The closed-loop feedback control module is used to receive the error information output by the adaptive calibration module and the temperature inversion module in real time, and feed the error information back to the laser module, the virtual reference channel construction module and the adaptive calibration module to achieve dynamic adaptive adjustment of the parameters of each module.

[0027] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0028] Through multi-scale dynamic modulation coding technology, it is possible to stably embed sinusoidal modulation and pulse coding features in the laser signal, thus forming a clear modulation signature in the frequency domain;

[0029] By combining digital matched filters with self-organizing maps and fuzzy clustering algorithms, accurate demodulation of reference fiber acquisition data and fusion of historical data are achieved, thus constructing a virtual reference channel reflecting the ideal reference state.

[0030] Through the extended Kalman filter combined with the fuzzy logic adaptive calibration algorithm, real-time recursive correction of the initial phase error between the optical fiber signal to be measured and the virtual reference signal is achieved, effectively compensating for errors caused by optical fiber transmission attenuation, laser drift and environmental noise, thereby greatly improving the accuracy of temperature inversion and the robustness of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the process of the present invention;

[0032] Figure 2 Schematic diagram of the virtual reference channel structure in the present invention;

[0033] Figure 3 Schematic diagram of adaptive calibration in the present invention;

[0034] Figure 4 This is a system structure diagram of the present invention. DETAILED DESCRIPTION

[0035] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0036] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise," "include," etc. used herein indicate the presence of the features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0037] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0038] An optical fiber temperature measurement method comprises the following steps:

[0039] The laser signal with multi-scale dynamic modulation coding is injected into the optical fiber to be tested and the reference optical fiber respectively, and the optical signal of each channel is collected and then subjected to fast Fourier transform to generate frequency domain data containing modulation coding characteristics;

[0040] The present invention proposes a fiber optic temperature measurement method. The core idea is to perform multi-scale dynamic modulation and coding processing on the laser signal, and inject the processed laser signal into the optical fiber to be measured and the reference optical fiber respectively, and then collect the two optical signals. The fast Fourier transform is realized by using digital signal processing technology to generate frequency domain data containing modulation and coding characteristics, thereby providing a reliable signal basis for the detection of temperature changes.

[0041] The overall process of this invention includes multiple steps, including laser signal modulation, optical signal injection, signal acquisition and analog-to-digital conversion, window function weighting, and fast Fourier transform (FFT). Each step is carefully designed to ensure accurate extraction of signal characteristics caused by temperature fluctuations even in complex environments. Specifically, a laser signal with multi-scale dynamic modulation and coding is first injected into the test fiber and the reference fiber, respectively. The test fiber is typically laid or wound in a temperature-sensitive area, while the reference fiber is installed in a temperature-controlled, low-vibration environment and serves as a reference for system calibration. During transmission, both optical fibers are affected by the external environment, mechanical stress, and system noise, resulting in subtle variations in optical signal parameters such as intensity and phase. To this end, during the acquisition phase, a high-speed optical receiver converts the two return optical signals into discrete analog signals, which are then converted to digital signals by a high-speed analog-to-digital converter (ADC). Subsequently, a digital signal processing module preprocesses the acquired time-domain data, including applying window function weighting to reduce spectral leakage. The preprocessed time-domain data is then converted into frequency-domain data using a fast Fourier transform (FFT) algorithm, yielding frequency-domain data containing modulation and coding characteristics. This frequency domain data not only retains the amplitude and phase information of the optical signal itself, but also forms obvious frequency peaks and pulse spectrum lines at specific frequencies due to the effect of modulation coding, which facilitates subsequent analysis.

[0042] The advantage of the whole process is that by pre-injecting specific modulation coding information at the laser end, the system forms stable identification features in the frequency domain. These identification features can be used to distinguish signal changes caused by ambient temperature changes from interference caused by system noise, laser drift and other factors. The present invention relies on high-speed digital signal processing and advanced Fourier transform technology. Even in long-distance transmission and high-noise environments, it can accurately extract weak signals caused by temperature changes, thereby achieving high-precision temperature monitoring. In short, the present invention provides a novel and efficient technical approach for temperature measurement by organically combining the modulation, transmission, acquisition and frequency domain conversion of optical signals. It also has good real-time and robustness, can be widely used in various industrial and environmental monitoring fields, and provides a solid data foundation for subsequent adaptive calibration and temperature inversion.

[0043] Preferably, in the multi-scale dynamic modulation coding step, a high-speed digital signal processor is used to perform sinusoidal wave modulation on the laser output signal, and the frequency of the sinusoidal wave modulation is set to not less than 1000 Hz and not more than 5000 Hz. At the same time, a pulse signal is superimposed by a dedicated pulse coding circuit, and the pulse width is set to not less than 5 microseconds and not more than 50 microseconds. The pulse repetition period is determined according to the temperature response characteristics of the environment to be measured, thereby generating a laser signal containing a fixed frequency component and a pulse coding component in the spectrum, and the laser signal is used to extract the temperature change mark in the subsequent frequency domain data.

[0044] In the multi-scale dynamic modulation coding step, the present invention uses a high-speed digital signal processor to implement a superposition of sinusoidal wave modulation and pulse coding on the laser output signal, so as to embed a specific modulation feature in the optical signal, thereby forming a fixed frequency component and a pulse coding component in the frequency domain, providing a unique identifier for the subsequent detection of temperature changes. In the specific implementation process, a high-speed digital signal processor is first used to generate a sinusoidal wave modulation signal, and its modulation frequency is set to , and satisfies , this frequency range has been experimentally verified to avoid overlap with common environmental noise while ensuring sufficient time resolution. The principle of sine wave modulation can be expressed as

[0045] ,

[0046] in, represents the average output amplitude of the laser, represents the modulation depth, is a time variable. This formula shows that after the laser signal is modulated by a sine wave, its output amplitude will change periodically and produce a corresponding frequency peak in the frequency domain.

[0047] At the same time, in order to enhance the multi-scale characteristics of the modulation signal, the system uses a dedicated pulse coding circuit to superimpose the pulse signal. The key parameters of pulse coding are pulse width and pulse repetition period, where the pulse width is set to no less than 5 and no more than 50 The pulse signal can be expressed as:

[0048] ,

[0049] In the above formula, represents the pulse amplitude, represents the pulse repetition period, represents the pulse width, and the function is a rectangular pulse function, which is defined as The value of the inner part is 1, and the value of the rest is 0. Pulse repetition period Reasonable determination is made based on the temperature response characteristics of the environment to be measured to balance time resolution and data redundancy.

[0050] After the sinusoidal wave modulation signal and the pulse coding signal are superimposed, the optical signal output by the laser can be expressed as:

[0051] ,

[0052] This composite signal exhibits distinct sinusoidal modulation peaks and periodic pulse-coded spectral lines in the frequency domain. The advantage of modulated coded signals lies in the fact that temperature variations during transmission of the optical signal through the fiber under test can cause subtle shifts in the phase and intensity of these modulated components. These shifts can be accurately identified through frequency domain analysis, reflecting changes in ambient temperature. In particular, since the reference fiber is in a relatively stable environment, its modulation coded signal exhibits minimal variations, making it an important benchmark for calibrating the signal from the fiber under test.

[0053] During implementation, a high-speed digital signal processor and dedicated pulse coding circuit must be coupled with a high-precision clock system to ensure that the phase and amplitude of the sinusoidal modulation and pulse coding are strictly controlled within the designed parameter range. Through offline calibration and online correction, the system can monitor and adjust the modulation parameters in real time, ensuring that the characteristics of the generated modulated and coded signal in the frequency domain are stable and reproducible. In summary, the multi-scale dynamic modulation and coding step provides a unique and reliable signal identification for the fiber-optic temperature measurement method, greatly improving the accuracy of subsequent temperature data extraction and the system's anti-interference ability.

[0054] Preferably, in the step of performing fast Fourier transform after collecting the optical signal, Hanning window weighting processing is first applied to the collected time domain data, and then Fourier transform is performed using a digital signal processing algorithm to generate frequency domain data, thereby ensuring that the modulation coding characteristics are retained in the frequency domain data.

[0055] In the fast Fourier transform step after collecting the optical signal, the optical signals returned by the test fiber and the reference fiber must first be converted into discrete digital time domain data through an analog-to-digital converter. In order to ensure that the frequency domain data can accurately reflect the characteristics of multi-scale dynamic modulation coding, Hanning window weighting is usually applied to the time domain data before performing the discrete Fourier transform. The mathematical expression of the Hanning window function is

[0056] ,

[0057] in, Represents the index of discrete sampling points, Indicates the length of the window function. This window function can make the sampled data exhibit a smooth attenuation effect at both ends of the data, reducing spectrum leakage caused by signal truncation.

[0058] For the original time domain data Multiply with the Hanning window function to obtain weighted data , whose expression is:

[0059] ,

[0060] Next, the weighted data is converted to the frequency domain using the fast Fourier transform algorithm to obtain the frequency domain representation ,Right now:

[0061] ,

[0062] In this process, the fast Fourier transform algorithm uses the divide-and-conquer strategy to efficiently calculate discrete data, and its output frequency domain data The fixed frequency components produced by the sine wave modulation and the periodic spectral lines introduced by the pulse coding are clearly shown. A clear spectrum peak is observed near the wavelength of the optical fiber under test, while pulse coding displays periodically distributed spectral lines at its harmonic positions. These modulation coding features provide a reliable anchor point for subsequent temperature change detection. When the optical signal is transmitted through the fiber under test, temperature changes will cause the amplitude or position of these spectral lines to shift slightly, thus reflecting the temperature change.

[0063] In order to ensure the accuracy and resolution of Fourier transform, the system requires the use of high sampling rate analog-to-digital converters in hardware design. The sampling rate is usually set to no less than 100 MHz, and the conversion bit depth is no less than 14 bits. At the same time, The number of points needs to be selected based on the actual sampling data length and the expected frequency domain resolution to ensure that the modulation coding features can be accurately distinguished in the frequency domain.

[0064] In the actual operation process, the digital signal processing module first stores the collected time domain data and then stores the data according to the set window function length. The Hanning window weights are calculated for each sampling point and then multiplied point by point with the original data to generate smooth weighted data. This weighted data is then fed into a fast Fourier transform algorithm, which uses a divide-and-conquer approach to obtain frequency domain data. To verify that the modulation coding characteristics are preserved, the system examines the spectral lines in the output frequency domain data. If the preset frequency components and pulse code components do not match expectations, it is determined that there is an anomaly in the sampling or modulation process, and appropriate corrective measures are initiated.

[0065] In summary, the steps of fast Fourier transform after collecting optical signals are achieved through Hanning window weighting and efficient The combination of algorithms realizes the accurate conversion from discrete time domain data to frequency domain data, ensuring that the frequency domain data contains stable and clear multi-scale modulation coding features, providing a solid data foundation for subsequent virtual reference construction, adaptive calibration and temperature inversion, thereby realizing high-precision fiber optic temperature measurement.

[0066] A virtual reference channel is constructed, wherein the reference fiber frequency domain data is demodulated and denoised, and the demodulated data is fused with pre-stored historical data to form virtual reference frequency domain data. The frequency domain data of the fiber under test are then compared with the virtual reference frequency domain data at the same sampling position, and adaptive calibration is performed using an adaptive calibration algorithm. The adaptive calibration algorithm uses an extended Kalman filter combined with fuzzy logic to correct the fiber under test signal to obtain calibrated phase data reflecting temperature changes.

[0067] The core steps of the present invention are to construct a virtual reference channel and use an adaptive calibration algorithm to correct the frequency domain data collected from the fiber under test, thereby obtaining highly accurate calibrated phase data that reflects temperature changes. Overall, this step first demodulates and suppresses noise in the frequency domain data collected from the reference fiber. The demodulated data is then fused with pre-stored historical reference data to construct virtual reference frequency domain data. Next, the frequency domain data collected from the fiber under test at the same sampling position is compared with the constructed virtual reference frequency domain data to calculate the initial phase error signal.

[0068] Utilizing an adaptive calibration algorithm based on an extended Kalman filter combined with fuzzy logic, the algorithm corrects the initial phase error signal through recursive prediction and updating. It also adaptively adjusts the process noise covariance and the measurement noise covariance online to gradually eliminate interference caused by ambient temperature changes, laser drift, and system noise. Ultimately, it outputs calibrated phase data that accurately reflects temperature changes. This core step plays a crucial role in the entire fiber-optic temperature measurement method, as only accurately calibrated phase data can be used as input to the temperature inversion module, enabling high-precision measurement of the temperature field distribution.

[0069] The entire process utilizes the complementary nature of dual-channel data. The reference fiber is in a stable environment, and its frequency domain data can serve as an ideal reference after demodulation and noise suppression. The data from the fiber under test contains a slight phase shift caused by temperature changes. By comparing the two at the same sampling position, the phase difference, i.e., the initial phase error signal, can be calculated. It can be expressed mathematically as:

[0070] ,

[0071] in, is the frequency domain data of the optical fiber to be tested, is the constructed virtual reference frequency domain data, Then, the extended Kalman filter combined with fuzzy logic is used to extract the phase. Recursive prediction and update are performed to achieve adaptive calibration. This process not only predicts the state variables based on the mathematical model, but also continuously adjusts the noise parameters through online feedback to ensure that the phase data output after calibration can accurately reflect temperature changes.

[0072] On the whole, this step fully integrates multiple links such as demodulation, data fusion, comparison and adaptive filtering, forming a complete closed-loop feedback control link, providing high-quality, low-noise basic data for subsequent temperature inversion, thereby greatly improving the accuracy and robustness of the fiber optic temperature measurement system. Through this core step, the system can not only effectively eliminate errors caused by laser drift and environmental noise, but also show extremely high sensitivity in detecting small phase shifts caused by temperature changes. In general, the present invention realizes the effective complementarity of the optical fiber to be tested and the reference optical fiber data by constructing a virtual reference channel and adaptive calibration, ensuring that high-precision temperature measurement results can be obtained in practical applications, and providing a reliable and efficient technical solution for the fields of industrial monitoring, environmental testing, etc.

[0073] Preferably, Figure 2 As shown, in the step of constructing a virtual reference channel, the frequency domain data collected by the reference optical fiber is demodulated and noise suppressed using a digital matched filter, and the demodulated data is fused with the pre-stored historical reference data using a self-organizing map algorithm combined with a fuzzy clustering algorithm to form virtual reference frequency domain data reflecting the ideal reference state.

[0074] In the step of constructing the virtual reference channel, in order to ensure that the frequency domain data collected by the reference fiber can effectively reflect the ideal reference state, the system first demodulates and suppresses the noise of the original frequency domain data collected by the reference fiber. Assume that the original frequency domain data collected by the reference fiber is ,in In order to extract the key components of the modulation coded signal, a digital matched filter is used to To process, the specific operation is to with preset modulation templates Perform convolution operation, the mathematical expression is:

[0075] ,

[0076] Here, the symbol represents the convolution operation, is the reference frequency domain data after demodulation and noise suppression. The digital matched filter can highlight the expected sinusoidal modulation and pulse coding components in the frequency domain by convolution with the modulation template, thereby effectively reducing the influence of system noise. Next, in order to construct the virtual reference frequency domain data under ideal conditions, the system converts the demodulated reference data The pre-stored historical reference data is fused. ,This data is a high-quality reference signal collected when the system is running stably, and its characteristics represent the ideal reference state.

[0077] The fusion process uses the self-organizing map algorithm (SOM) combined with the fuzzy clustering algorithm. This algorithm can cluster and summarize the current demodulated data and historical data in a high-dimensional feature space, thereby extracting the most representative reference features. The fusion process can be expressed as:

[0078] ,

[0079] in, Represents the processing of the self-organizing map algorithm, the output This is the constructed virtual reference frequency domain data. This method not only incorporates the effective modulation and coding features of the current reference fiber data, but also incorporates the stable feature information from the historical data, thereby forming a benchmark data with a high signal-to-noise ratio in the frequency domain that can reflect the ideal reference state.

[0080] In actual operation, in order to ensure the data fusion effect, the parameters of the self-organizing mapping algorithm (such as the number of cluster centers, learning rate, etc.) need to be determined in advance during the offline calibration stage and dynamically adjusted according to data changes during online operation. This fusion process has good robustness and adaptability, and can output a stable virtual reference signal under various environmental interferences, providing an accurate benchmark for subsequent adaptive calibration. In addition, the digital matched filter needs to fully consider the sampling rate and modulation template matching during the processing process to ensure that the modulation coding features are obvious in the convolution operation results, while suppressing non-target noise. Through precise parameter settings and strict hardware selection, this step achieves a smooth transition from original reference data to virtual reference data, and its output data It is a crucial calibration reference in the entire fiber optic temperature measurement system.

[0081] The step of constructing a virtual reference channel plays a critical role in the system because it provides an ideal comparison standard for the correction of the optical fiber data to be measured. By fusing the current demodulated data with historical reference data, the system can automatically compensate for errors caused by signal attenuation, laser drift, and environmental noise during optical fiber transmission, making the virtual reference signal closer to the theoretical ideal state. Ultimately, the implementation of this step not only improves the stability of the reference signal, but also provides accurate input for the subsequent adaptive calibration module, thereby greatly improving the overall accuracy and robustness of the temperature measurement. The implementation of the entire process requires the hardware to have high-speed digital signal processing capabilities and high-precision matched filters. At the same time, the software algorithm needs to implement efficient calculations of self-organizing maps and fuzzy clustering algorithms to ensure that the virtual reference frequency domain data can be output in real time during online operation.

[0082] Preferably, Figure 3 As shown in the figure, in the adaptive calibration step, the frequency domain data of the optical fiber to be tested is compared with the virtual reference frequency domain data at the same sampling position to calculate the initial phase error signal, and the extended Kalman filter combined with fuzzy logic is used to recursively predict and update the initial phase error signal. At the same time, the process noise covariance and the measurement noise covariance are adaptively adjusted online to obtain calibrated phase data that accurately reflects temperature changes.

[0083] In the adaptive calibration step, the system compares the frequency domain data collected by the optical fiber under test with the constructed virtual reference frequency domain data at the same sampling position to calculate the initial phase error signal, and uses the extended Kalman filter combined with fuzzy logic to recursively predict and update the initial phase error signal to obtain the calibrated phase data that accurately reflects the temperature change. Specifically, suppose the frequency domain data of the optical fiber under test is The virtual reference frequency domain data is At the same sampling position, the phases of the two are recorded as and The initial phase error signal is defined as:

[0084] ,

[0085] The initial phase error signal reflects the difference between the optical fiber signal to be measured and the ideal reference signal due to temperature changes. In order to correct this error signal, the system uses an extended Kalman filter, which recursively estimates the phase error based on the state space model. Let the state variable Indicates sampling point The true phase error at , the state transfer equation can be written as:

[0086] ,

[0087] in, is the state transition matrix, is the process noise; the measurement model is:

[0088] ,

[0089] in, is the measured initial phase error signal, is the observation matrix, is the measurement noise. The prediction steps of the extended Kalman filter are:

[0090] ,

[0091] ,

[0092] in, Represents the predicted state, is the forecast error covariance, is the process noise covariance matrix. The update steps are:

[0093] ,

[0094] ,

[0095] ,

[0096] in, is the Kalman gain, is the measurement noise covariance matrix, In order to adapt to the changes in the environment and system status during actual operation, the present invention introduces a fuzzy logic mechanism into the extended Kalman filter to calculate the noise covariance matrix. and Perform online adaptive adjustment. Specifically, when the filter residual increases, the fuzzy rule will automatically adjust to increase the uncertainty of the forecast or adjust the The value of t is used to reduce the dependence on the current measurement, so that the filter can respond quickly to environmental changes. After multiple predictions and update iterations, the final output state estimate That is the calibrated phase data, which can accurately reflect the phase offset caused by temperature changes.

[0097] In a preferred embodiment, the adaptive calibration step also includes continuously monitoring the calibration results using a feedback mechanism during the online processing process, and further iteratively optimizing the filter parameters using an online minimum mean square error algorithm. Through this closed-loop control, the system can adjust the various parameters of the adaptive calibration algorithm in real time, ensuring that the calibrated phase data maintains high precision and low noise during long-term operation. The successful implementation of this step depends on the support of a high-precision digital signal processor and sufficient offline calibration work. The initial parameters are determined through experiments and then continuously updated through online feedback. In practical applications, the response speed and accuracy of the filter can be finely controlled by setting the sampling period, update frequency, and error threshold, thereby achieving accurate capture and correction of the phase signal due to temperature changes. Overall, the adaptive calibration step, based on the stable reference data provided by the virtual reference channel, uses the extended Kalman filter and fuzzy logic to achieve fine correction of the measured signal, providing extremely reliable data input for subsequent temperature inversion, so that the entire fiber optic temperature measurement system can still output high-precision temperature distribution data in complex environments.

[0098] The temperature inversion of the calibrated phase data is performed based on the pre-established temperature inversion model, and the preliminary temperature data is optimized using multidimensional correlation analysis to output high-precision temperature distribution data.

[0099] The present invention aims to achieve high-precision real-time measurement of the temperature field distribution in optical fiber temperature measurement. The core idea is to perform temperature inversion on the calibrated phase data after adaptive calibration based on a pre-established temperature inversion model, and to further optimize the preliminary temperature data in combination with multidimensional correlation analysis, thereby outputting high-precision temperature distribution data reflecting temperature changes.

[0100] The present invention generally digitizes the optical signals collected from the test fiber and the reference fiber to obtain frequency domain data containing modulation coding characteristics. A virtual reference channel structure and an adaptive calibration module are then used to eliminate system noise, laser drift, and external environmental interference, ensuring that the output calibrated phase data accurately reflects temperature changes in the test area. Subsequently, based on a temperature inversion model obtained through offline calibration, the calibrated phase data is converted into temperature values ​​to obtain preliminary temperature data. This is followed by a cross-spectral analysis of the preliminary temperature data and the modulation coding characteristic data using multidimensional correlation analysis to extract the nonlinear correlation between the two. The preliminary temperature data is then robustly optimized using a random sampling consistency algorithm to ultimately obtain corrected temperature data.

[0101] In addition, in order to ensure the accuracy and stability of the system during long-term operation, the present invention also adopts a closed-loop feedback control mechanism to feed back the filter residuals generated during the adaptive calibration process to the multi-scale dynamic modulation coding and virtual reference channel construction module of the front end in real time, and uses the online minimum mean square error algorithm to implement dynamic adaptive adjustment of relevant parameters. Overall, the present invention achieves high-precision capture and accurate inversion of small phase changes caused by temperature through the organic combination of constructing virtual reference channels, adaptive calibration, temperature inversion and multi-dimensional correlation analysis and closed-loop feedback mechanism. This solution makes full use of the complementary information between the optical fiber to be tested and the reference optical fiber during the transmission process, which not only ensures that the modulation and coding characteristics of the signal in the frequency domain are completely preserved, but also effectively eliminates interference factors inside and outside the system through adaptive calibration, so that the temperature inversion result has high accuracy and robustness.

[0102] This invention enables stable output of accurate temperature distribution data, even in complex environmental noise environments or over long fiber transmission distances, providing a reliable and efficient solution for applications such as industrial process monitoring, environmental monitoring, and safety warnings. The overall goal of this invention is to achieve high-precision conversion from raw calibrated phase data to final temperature distribution data. Each step achieves optimal performance through offline calibration and online adaptive adjustment. Advanced technologies such as multidimensional data fusion and cross-spectral analysis ensure accurate recovery and real-time monitoring of temperature information, forming a complete, closed-loop feedback temperature measurement system.

[0103] Preferably, the temperature inversion step adopts a temperature inversion model established based on offline calibration data, and the temperature inversion model adopts quadratic polynomial fitting to convert the calibrated phase data into temperature values, thereby generating preliminary temperature data.

[0104] In the temperature inversion step, the system uses a temperature inversion model built based on offline calibration data to convert the calibrated phase data into temperature values, thereby generating preliminary temperature data. Specifically, during the offline calibration phase, the experimenter collects calibrated phase data under known temperature conditions to establish a mapping relationship between temperature and phase change. This mapping relationship is usually modeled using a quadratic polynomial fitting method, and its mathematical expression is:

[0105] ,

[0106] in, Indicates the temperature value, It represents the phase shift in the calibrated phase data that reflects the temperature change. 、 as well as are the fitting coefficients obtained through offline calibration. In practical applications, the system inputs the calibrated phase data into the temperature inversion module. This module calculates according to the aforementioned quadratic polynomial model, converting the phase offset of each sampling point into a corresponding temperature value, thereby generating preliminary temperature data. To ensure the accuracy and adaptability of the model, sufficient sample data should be collected during offline calibration to cover the expected temperature range. Furthermore, during system operation, regular calibration or real-time feedback correction can be used to update the coefficients of the temperature inversion model to ensure that the inversion results always reflect actual temperature changes.

[0107] The operational details of this step include: first, obtaining the calibrated phase data through the data acquisition module, and the data is in the form of a discrete phase value sequence; second, the phase data is operated according to the preset quadratic polynomial model to calculate the temperature value of each sampling point; third, the output preliminary temperature data is the temperature value sequence corresponding to each sampling point, which can be directly used for subsequent optimization processing. In this embodiment, if the calibrated phase data of a certain sampling point is , then the corresponding temperature value The formula Calculated.

[0108] In actual application, the digital signal processing module performs the quadratic fitting operation through a built-in algorithm and outputs the temperature data in real time on a high-speed processor. The advantage of using the quadratic polynomial fitting method in the temperature inversion model is that it has low computational complexity and can better describe the nonlinear relationship between temperature and phase change in most cases. In addition, the model has a simple structure and is convenient for offline calibration and online updating, thereby ensuring the inversion accuracy in actual temperature measurement and facilitating hardware implementation. Through this temperature inversion step, the preliminary temperature data reflects the temperature information in the calibrated phase data, but due to the possible measurement errors and noise in the actual environment, the preliminary temperature data may still have a certain degree of uncertainty. Therefore, this step lays the foundation for the subsequent use of multidimensional correlation analysis to further optimize the preliminary temperature data, ensuring that the final output temperature data has high accuracy and low noise.

[0109] Preferably, in the multidimensional correlation analysis step, cross-spectrum analysis is performed on the preliminary temperature data and the modulation coding feature data, the nonlinear correlation between the two is extracted through normalization processing and correlation coefficient calculation, and the preliminary temperature data is robustly optimized using a random sampling consistency algorithm to obtain the corrected temperature data.

[0110] In the multidimensional correlation analysis step, the system further performs a joint analysis on the preliminary temperature data and the modulation coding feature data to extract the nonlinear correlation between the two, and uses the random sampling consistency algorithm to perform robust optimization on the preliminary temperature data to obtain the corrected temperature data. First, the preliminary temperature data is generated by the temperature inversion step, which represents the temperature estimate obtained after quadratic polynomial fitting. However, due to factors such as system noise, optical fiber transmission loss and environmental interference, there may still be certain random errors in the preliminary temperature data. In order to further eliminate these interference factors, the system performs cross-spectral analysis on the preliminary temperature data and the modulation coding feature data obtained in the multi-scale dynamic modulation coding step. Cross-spectral analysis is a method for detecting the correlation between two sets of signals in the frequency domain. By normalizing the preliminary temperature data and the modulation coding feature data and calculating the correlation coefficient, the nonlinear correlation information between the two can be extracted, thereby effectively identifying the frequency domain features related to temperature changes. Assume that the preliminary temperature data is , the modulation coding characteristic data is , then the cross spectrum function can be expressed as:

[0111] ,

[0112] in, represents the complex conjugate of the modulation coding feature data, Represents the amplitude. By calculating the cross-spectral function, a correlation coefficient can be obtained, which reflects the correlation between the preliminary temperature data and the modulation coding feature data at each frequency domain sampling point. Subsequently, the system uses a random sampling consistency algorithm to perform robust optimization on the preliminary temperature data. This algorithm eliminates outliers and noise interference by searching for temperature data that is consistent with the cross-spectral analysis results in a large number of random samples, thereby optimizing the accuracy of the preliminary temperature data. The core of the random sampling consistency algorithm is to screen and cluster the data set through multiple iterations to obtain a set of temperature data with high cohesion, which is recorded as , which is the corrected temperature data after robust optimization.

[0113] The combination of multidimensional correlation analysis and robust optimization offers significant advantages. It leverages the stability of modulation and coding features in the frequency domain while revealing nonlinear information in temperature data through cross-spectral analysis, effectively reducing noise interference and random errors in complex environments. In practice, this step requires the digital signal processing module to possess high-speed computing capabilities and to preprocess, normalize, and calculate correlation coefficients for the collected data. For example, if the correlation coefficient between the preliminary temperature data and the modulation and coding feature data at a certain frequency domain sampling point reaches 0.85, the system considers the temperature data at that point to be relatively reliable. If the correlation coefficient is lower, the system will eliminate the data at that point through a random sampling consistency algorithm. Ultimately, through multidimensional correlation analysis and robust optimization, the system outputs corrected temperature data that not only accurately reflects temperature changes but also exhibits high robustness and consistency, providing high-quality temperature information for the entire temperature measurement system. The implementation of this step requires careful debugging of the normalization method, correlation coefficient calculation formula, and parameters of the random sampling consistency algorithm involved in the cross-spectrum analysis to ensure stable results under various measurement conditions, thereby providing a solid data foundation for subsequent temperature data feedback and overall closed-loop control of the system.

[0114] Preferably, the calibrated phase data is monitored in real time through a closed-loop feedback system, which transmits the filtering residuals generated during the calibration process to the multi-scale dynamic modulation coding and virtual reference channel construction steps, and uses an online minimum mean square error algorithm to perform dynamic adaptive adjustment of the modulation parameters and virtual reference construction parameters.

[0115] In the subsequent process of the temperature inversion step, in order to further ensure the long-term stability and high precision of the temperature measurement system, the present invention designs a closed-loop feedback system to monitor the calibrated phase data in real time, and pass the filter residuals generated in the calibration process to the multi-scale dynamic modulation coding and virtual reference channel construction steps, thereby realizing online adaptive adjustment of parameters. Specifically, when the system performs temperature inversion, it uses a pre-established temperature inversion model to convert the calibrated phase data into preliminary temperature data, and then outputs the corrected temperature data after multi-dimensional correlation analysis optimization. In this process, any errors caused by changes in the external environment or system drift will appear as residual signals in the filter. Suppose the residual signal output by the filter is , this residual signal reflects the error of the extended Kalman filter during the adaptive calibration process. When the residual signal exceeds the preset threshold, the closed-loop feedback system automatically transmits this information to the multi-scale dynamic modulation coding module and virtual reference channel construction module at the front end. The specific feedback mechanism is based on an online minimum mean square error algorithm. By calculating the error between the current system output temperature and the auxiliary temperature reference data in real time, the laser signal modulation parameters and virtual reference construction parameters are adjusted. The feedback control process can be expressed as follows:

[0116] ,

[0117] in, represents the parameter to be adjusted (e.g., sine wave modulation depth, pulse repetition period, or cluster center parameter for virtual reference construction), Indicates the current temperature error, is the learning rate. Through this iterative adjustment mechanism, the system can optimize the front-end parameter settings in real time, ensuring an optimal match between the modulation coding characteristics of the measured fiber frequency domain data and the reference channel data. This enables the extended Kalman filter to more accurately correct phase errors during the adaptive calibration process.

[0118] In practical applications, closed-loop feedback systems require data acquisition, processing, and feedback control to be completed within milliseconds to ensure rapid response to environmental changes. This requires hardware implementation of a high-speed data processor and real-time control unit, while software implementation requires an online minimum mean square error (MMSE) algorithm and a dynamic gradient descent adjustment mechanism. The adjusted parameters are immediately applied to laser modulation and virtual reference data construction, forming a dynamically adaptive closed-loop control system. This ensures that the entire temperature measurement process maintains high accuracy and stability in the output temperature data despite variations in laser drift, ambient noise, or fiber transmission loss. Overall, this closed-loop feedback and online adaptive adjustment mechanism, through real-time monitoring of filter residuals and dynamic correction of front-end parameters using an online MMSE algorithm, effectively mitigates the drift and error accumulation that can occur in traditional open-loop systems over long periods of operation, providing the entire fiber-optic temperature measurement system with extremely high stability and robustness. Experimental verification demonstrates that this mechanism operates stably under a wide range of operating conditions, ensuring that the temperature inversion results remain consistent with the actual ambient temperature, significantly improving the system's practical performance.

[0119] Preferably, the temperature inversion step further includes iteratively adjusting the temperature inversion model parameters using an online minimum mean square error algorithm, wherein the online minimum mean square error algorithm uses the error between the temperature value obtained by converting the auxiliary temperature reference data and the calibrated phase data as feedback to reduce the error and improve the accuracy of the temperature inversion.

[0120] The temperature inversion step further iteratively adjusts the temperature inversion model parameters using an online minimum mean square error algorithm to reduce the error between the auxiliary temperature reference data and the temperature values ​​converted from the calibrated phase data, thereby improving the accuracy of the temperature inversion. Specifically, in the preliminary temperature inversion step, the system uses a temperature inversion model established based on offline calibration data and a quadratic polynomial fitting method to convert the calibrated phase data into preliminary temperature data. The model can be expressed as

[0121] ,

[0122] in, Indicates the temperature value, It represents the phase shift in the calibrated phase data that reflects the temperature change. 、 as well as is the fitting coefficient determined during offline calibration. However, during actual operation, due to the influence of the external environment, equipment aging or other non-ideal factors, the temperature inversion model may have certain deviations. To this end, the system introduces an online minimum mean square error algorithm to iteratively adjust the temperature inversion model parameters. The online minimum mean square error algorithm continuously calculates the error between the auxiliary temperature reference data and the temperature value converted from the calibrated phase data, and then uses this error to perform a gradient descent update on the model parameters. The formula can be expressed as:

[0123] ,

[0124] in, Indicates the first coefficients, represents the mean square value of the temperature error, is the learning rate. This online adjustment mechanism enables the temperature inversion model to adaptively correct parameter deviations, ensuring that the error between the converted temperature value and the actual measured value of the auxiliary temperature reference data is as small as possible. During implementation, the system will periodically or in real time obtain the auxiliary temperature reference data, and compare it with the temperature value currently calculated by the model. The calculated error is used as feedback input into the online minimum mean square error algorithm, thereby continuously updating the model parameters. The specific operation steps include: first, collecting the calibrated phase data and calculating the preliminary temperature data using the initial temperature inversion model; second, obtaining the auxiliary temperature reference data and calculating the error between the preliminary temperature data and the reference data; third, using the error calculation results, updating the coefficients of the temperature inversion model through the gradient descent method; finally, using the updated model to recalculate the temperature value, and continue to iterate until the error converges to the preset allowable range.

[0125] In practical applications, this online minimum mean square error algorithm requires the system to possess efficient data acquisition and processing capabilities, while ensuring a sufficiently high update frequency to cope with environmental changes. Practical implementations have demonstrated that online iterative adjustment not only significantly reduces the systematic error of temperature inversion but also maintains the stability and accuracy of temperature data output during long-term monitoring. This mechanism enables the temperature inversion model to dynamically adapt to changes in external conditions, forming a closed-loop feedback loop and working in conjunction with other modules (such as multi-scale dynamic modulation coding and virtual reference channel construction), thereby further improving the performance of the overall fiber-optic temperature measurement system. Through this online iterative parameter adjustment method, the system achieves high-precision conversion from calibrated phase data to temperature values, providing real-time adaptive correction for the entire temperature measurement process, ensuring that the temperature inversion results maintain high accuracy and low error in a variety of complex environments.

[0126] like Figure 4 As shown, an optical fiber temperature measurement system is used to implement the optical fiber temperature measurement method, the system comprising:

[0127] The system is equipped with a laser module, which is used to output laser signals that have undergone multi-scale dynamic modulation and coding. This module uses a high-coherence, frequency-adjustable continuous wave laser, which is precisely controlled to output laser signals that have undergone multi-scale dynamic modulation and coding. In the hardware design, the laser drive circuit works in conjunction with the digital signal processor, and the built-in modulation circuit realizes the superposition of sinusoidal wave modulation and pulse coding. Specifically, the laser module is embedded with a high-speed digital signal processor, which generates the modulation signal. The frequency of the sinusoidal wave modulation part is set to meet 1000 Hz ≤ ≤5000 Hz, and a dedicated pulse encoding circuit generates pulse signals with pulse widths between 5 and 50 microseconds. The pulse repetition period is set based on the temperature response characteristics of the measured environment. The output laser signal exhibits both sinusoidal modulation and pulse encoding characteristics in the time domain, and can be displayed as a fixed frequency component and periodic pulse spectrum in the frequency domain, providing clear identification information for subsequent signal processing.

[0128] The optical path distribution module is used to divide the modulated laser signal equally and inject it into the optical fiber to be tested and the reference optical fiber respectively. The optical fiber to be tested is used to transmit to the object to be tested, and the reference optical fiber is placed in a temperature-controlled and low-vibration environment. The optical path distribution module is responsible for dividing the modulated laser signal output by the laser module equally and injecting it into two different optical fiber channels respectively. The module uses a low-loss, high-precision optical splitter to divide the modulated signal into two parts: one part is injected into the optical fiber to be tested, which is directly transmitted to the area where the object to be tested is located and is used to collect optical signal changes caused by ambient temperature changes; the other part is injected into the reference optical fiber, which is placed in a temperature-controlled and low-vibration environment and is used to collect stable reference optical signals to provide a system calibration benchmark. When designing the optical path distribution module, it is necessary to ensure that the two optical signals have the same optical power and phase stability. At the same time, strict requirements are placed on the selection of components such as optical fiber connectors and couplers to avoid additional interference due to component errors.

[0129] The signal detection module collects the return optical signals from the test fiber and the reference fiber and converts them into digital time-domain data. The system's signal detection module utilizes a high-speed optical receiver and a high-precision analog-to-digital converter (ADC) to collect the return optical signals from the test fiber and the reference fiber. The optical receiver converts the optical signal into an electrical signal, which is then amplified by a low-noise amplifier. The ADC then converts the analog signal into digital time-domain data. Hardware-wise, this module typically requires an ADC sampling rate of at least 100 MHz and a resolution of at least 14 bits to ensure the collected digital data has sufficient dynamic range and resolution to fully capture subtle variations in the modulated coded signal. The signal detection module outputs two channels of digital time-domain data, providing the raw input for subsequent digital signal processing.

[0130] The digital signal processing module is used to perform a fast Fourier transform (FFT) on the digital time-domain data, generating frequency-domain data containing modulation and coding features after Hanning window weighting. The digital signal processing module is responsible for performing a fast Fourier transform (FFT) on the collected digital time-domain data. This module integrates a high-speed digital signal processor (DSP) or field-programmable gate array (FPGA) to enable real-time data processing. This module first applies a Hanning window weighting to the time-domain data.

[0131] The virtual reference channel construction module demodulates and suppresses noise in the frequency domain data collected by the reference fiber, fusing the demodulated data with pre-stored historical reference data to form virtual reference frequency domain data reflecting the ideal reference state. This module employs a dedicated digital signal processor and matched filter in hardware.

[0132] The adaptive calibration module compares the frequency domain data of the fiber under test with the virtual reference frequency domain data at the same sampling location. It uses an extended Kalman filter combined with fuzzy logic to recursively predict and update the initial phase error signal. It also performs online adaptive adjustment of the process noise covariance and the measurement noise covariance, thereby outputting calibrated phase data that reflects temperature changes. The adaptive calibration module primarily compares the frequency domain data of the fiber under test with the virtual reference frequency domain data at the same sampling location. It uses an adaptive calibration algorithm to recursively predict and update the initial phase error signal, thereby outputting calibrated phase data that reflects temperature changes. This module typically consists of a high-performance embedded processor with a built-in extended Kalman filter and fuzzy logic controller.

[0133] The temperature inversion module converts the calibrated phase data into temperature values ​​based on a pre-established temperature inversion model and optimizes the preliminary temperature data using multidimensional correlation analysis to output high-precision temperature distribution data. The temperature inversion module performs temperature inversion on the calibrated phase data based on the pre-established temperature inversion model and optimizes the preliminary temperature data using multidimensional correlation analysis, thereby outputting high-precision temperature distribution data. In terms of hardware, this module consists of a dedicated embedded processor and memory, and its temperature inversion model typically uses a quadratic polynomial fitting method. A digital signal processor substitutes the calibrated phase data into the above model to calculate preliminary temperature data. The temperature inversion module then uses multidimensional correlation analysis to perform cross-spectral analysis on the preliminary temperature data and modulation coding feature data. Normalization and correlation coefficient calculation are used to extract the nonlinear correlation between the two. A random sampling consistency algorithm is then used to robustly optimize the preliminary temperature data, ultimately outputting corrected temperature data. The hardware implementation of this module requires high computing power and memory bandwidth to process large amounts of data in real time. Built-in algorithms continuously update and optimize model parameters to ensure high accuracy and stability of temperature data.

[0134] The closed-loop feedback control module is used to receive the error information output by the adaptive calibration module and the temperature inversion module in real time, and to feed back the error information to the laser module, the virtual reference channel construction module, and the adaptive calibration module to achieve dynamic adaptive adjustment of the parameters of each module. The closed-loop feedback control module is used to achieve dynamic adaptive adjustment of system parameters. This module receives the error information output by the adaptive calibration module and the temperature inversion module in real time, such as the filter residual and temperature error, and then calculates the adjustment amount through the online minimum mean square error algorithm. The closed-loop feedback control module transmits the calculated adjustment amount to the laser module, the virtual reference channel construction module, and the adaptive calibration module in real time, thereby achieving coordinated online adaptive adjustment of the parameters of each module. In terms of hardware, this module is usually composed of a low-latency controller and a high-speed data communication interface to ensure that the feedback information is transmitted within milliseconds and applied to the front-end parameter adjustment, so that the entire system maintains high precision and stability when the environment and system status change.

[0135] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware.

[0136] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included within the scope of the claims of the present application.

Claims

1. A fiber optic temperature measurement method, characterized in that: The steps include: The laser signal with multi-scale dynamic modulation coding is injected into the optical fiber to be tested and the reference optical fiber respectively, and the optical signal of each channel is collected and then subjected to fast Fourier transform to generate frequency domain data containing modulation coding characteristics; A virtual reference channel is constructed, wherein the reference fiber frequency domain data is demodulated and denoised, and the demodulated data is fused with pre-stored historical data to form virtual reference frequency domain data. The frequency domain data of the fiber under test are then compared with the virtual reference frequency domain data at the same sampling position, and adaptive calibration is performed using an adaptive calibration algorithm. The adaptive calibration algorithm uses an extended Kalman filter combined with fuzzy logic to correct the fiber under test signal to obtain calibrated phase data reflecting temperature changes. The temperature inversion of the calibrated phase data is performed based on the pre-established temperature inversion model, and the preliminary temperature data is optimized using multidimensional correlation analysis to output high-precision temperature distribution data.

2. The optical fiber temperature measurement method according to claim 1, characterized in that: In the multi-scale dynamic modulation coding step, a high-speed digital signal processor is used to perform sinusoidal wave modulation on the laser output signal. The frequency of the sinusoidal wave modulation is set to not less than 1000 Hz and not more than 5000 Hz. At the same time, the pulse signal is superimposed through a dedicated pulse coding circuit. The pulse width is set to not less than 5 microseconds and not more than 50 microseconds. The pulse repetition period is determined according to the temperature response characteristics of the environment to be measured, thereby generating a laser signal containing a fixed frequency component and a pulse coding component in the spectrum. The laser signal is used to extract temperature change identification in subsequent frequency domain data.

3. The optical fiber temperature measurement method according to claim 1, characterized in that: After collecting the optical signal, the fast Fourier transform step first applies Hanning window weighting processing to the collected time domain data, and then uses a digital signal processing algorithm to perform Fourier transform to generate frequency domain data, thereby ensuring that the modulation coding characteristics are retained in the frequency domain data.

4. The optical fiber temperature measurement method according to claim 1, characterized in that: In the step of constructing a virtual reference channel, the frequency domain data collected by the reference optical fiber is demodulated and noise suppressed using a digital matched filter, and the demodulated data is fused with the pre-stored historical reference data using a self-organizing map algorithm combined with a fuzzy clustering algorithm to form virtual reference frequency domain data reflecting the ideal reference state.

5. The optical fiber temperature measurement method according to claim 1, characterized in that: In the adaptive calibration step, the frequency domain data of the optical fiber to be tested is compared with the virtual reference frequency domain data at the same sampling position to calculate the initial phase error signal. The extended Kalman filter combined with fuzzy logic is used to recursively predict and update the initial phase error signal. At the same time, the process noise covariance and the measurement noise covariance are adaptively adjusted online to obtain calibrated phase data that accurately reflects temperature changes.

6. The optical fiber temperature measurement method according to claim 1, characterized in that: In the temperature inversion step, a temperature inversion model established based on offline calibration data is used. The temperature inversion model uses quadratic polynomial fitting to convert the calibrated phase data into temperature values, thereby generating preliminary temperature data.

7. The optical fiber temperature measurement method according to claim 1, characterized in that: In the multidimensional correlation analysis step, cross-spectral analysis is performed on the preliminary temperature data and the modulation coding feature data. The nonlinear correlation between the two is extracted through normalization processing and correlation coefficient calculation. The preliminary temperature data is robustly optimized using a random sampling consistency algorithm to obtain the corrected temperature data.

8. The optical fiber temperature measurement method according to claim 1, characterized in that: The calibrated phase data is monitored in real time through a closed-loop feedback system. The closed-loop feedback system transmits the filtering residuals generated during the calibration process to the multi-scale dynamic modulation coding and virtual reference channel construction steps, and uses an online minimum mean square error algorithm to dynamically and adaptively adjust the modulation parameters and virtual reference construction parameters.

9. The optical fiber temperature measurement method according to claim 1, characterized in that: The temperature inversion step further includes iteratively adjusting the temperature inversion model parameters using an online minimum mean square error algorithm, wherein the online minimum mean square error algorithm uses the error between the temperature value converted from the auxiliary temperature reference data and the calibrated phase data as feedback to reduce the error and improve the accuracy of the temperature inversion.

10. An optical fiber temperature measurement system, used to implement the optical fiber temperature measurement method according to any one of claims 1 to 9, characterized in that: The system includes: A laser module, configured to output a laser signal that has undergone multi-scale dynamic modulation encoding; An optical path distribution module is used to equally divide the modulated laser signal and inject it into the optical fiber under test and the reference optical fiber respectively. The optical fiber under test is used to transmit to the object under test, and the reference optical fiber is placed in a temperature-controlled and low-vibration environment; A signal detection module is used to collect the return optical signals of the optical fiber to be tested and the reference optical fiber, and convert the optical signals into digital time domain data; A digital signal processing module, configured to perform a fast Fourier transform on the digital time domain data, and generate frequency domain data containing modulation coding features after Hanning window weighting processing; A virtual reference channel construction module is used to demodulate and suppress noise on the frequency domain data collected by the reference optical fiber, and fuse the demodulated data with pre-stored historical reference data to form virtual reference frequency domain data reflecting an ideal reference state; The adaptive calibration module is used to compare the frequency domain data of the optical fiber to be tested with the virtual reference frequency domain data at the same sampling position, recursively predict and update the initial phase error signal using an extended Kalman filter combined with fuzzy logic, and perform online adaptive adjustment of the process noise covariance and the measurement noise covariance to output calibrated phase data reflecting temperature changes; a temperature inversion module, configured to convert the calibrated phase data into temperature values ​​based on a pre-established temperature inversion model, and optimize the preliminary temperature data using multidimensional correlation analysis to output high-precision temperature distribution data; The closed-loop feedback control module is used to receive the error information output by the adaptive calibration module and the temperature inversion module in real time, and feed the error information back to the laser module, the virtual reference channel construction module and the adaptive calibration module to achieve dynamic adaptive adjustment of the parameters of each module.

Images