Fiber grating-based stress monitoring and temperature measurement method and system

Abstract

The application relates to the technical field of optical fiber sensing, and discloses a stress monitoring and temperature measuring method and system based on an optical fiber grating. The method obtains identification information of a target optical fiber grating sensor, queries and determines a historical stress temperature data set in a historical storage unit, performs grouping operation on the historical stress temperature data set to generate a plurality of grouped monitoring data sets, extracts wavelength shift features for each grouped monitoring data set to obtain a plurality of characteristic core values and characteristic influence ranges, establishes a plurality of analysis units by using the characteristic influence ranges, applies the analysis units to process a real-time captured optical signal sequence to generate a plurality of analysis characteristic sets, integrates the analysis characteristic sets to obtain an integrated characteristic result, and then derives a stress measurement value and a temperature measurement value based on the result. The method improves the effectiveness and reliability of the optical fiber grating in stress and temperature monitoring, and is suitable for related measurement scenes in the fields of engineering structure health monitoring and industrial process control.

Description

Technical Field

[0001] This application relates to the field of fiber optic sensing technology, and in particular to a stress monitoring and temperature measurement method and system based on fiber Bragg gratings. Background Technology

[0002] In fields such as structural health monitoring and industrial process control, accurate measurement of environmental stress and temperature has significant practical application value. Fiber Bragg grating (FBG) sensors, with their characteristics of electromagnetic interference resistance, corrosion resistance, small size, and distributed measurement capability, have become important tools for stress and temperature monitoring. Their working principle is based on the wavelength shift that occurs in fiber Bragg gratings under external stress or temperature; by detecting this wavelength shift, the corresponding stress or temperature change can be deduced.

[0003] Existing fiber Bragg grating-based monitoring methods still have many limitations in practical applications. On the one hand, the wavelength shift of fiber Bragg gratings is simultaneously affected by the combined effects of stress and temperature; the measurement signal from a single sensor cannot directly distinguish the independent effects of stress and temperature, easily leading to measurement errors. On the other hand, traditional methods often rely solely on real-time acquired signals for analysis, neglecting the operating condition characteristics and patterns contained in historical monitoring data. The response characteristics of fiber Bragg gratings vary under different operating conditions; the lack of effective utilization of historical data results in insufficient adaptability of the analysis model, making it difficult to maintain stable measurement accuracy in complex and changing environments.

[0004] Current data processing methods lack precision in grouping and feature extraction of monitoring data. Most methods treat historical data as a whole for general analysis, failing to consider the differences in data across different time periods and environmental conditions. This results in weakly targeted feature extraction, unable to accurately reflect stress and temperature variation patterns under specific operating conditions. Furthermore, when processing real-time signals, there is a lack of effective correlation with historical features, and the construction of real-time signal analysis models does not fully incorporate the feature ranges from historical data, making it difficult to guarantee the reliability of real-time measurement results. These problems limit the further application of fiber Bragg grating sensors in high-precision monitoring scenarios, necessitating a monitoring method that can effectively integrate historical data, accurately extract features, and optimize real-time signal processing. Summary of the Invention

[0005] The purpose of this invention is to provide a stress monitoring and temperature measurement method based on fiber Bragg gratings to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a stress monitoring and temperature measurement method based on fiber Bragg gratings, the method comprising:

[0007] The identification information of the target fiber Bragg grating sensor is obtained, and a query operation is performed in the historical storage unit based on the identification information to determine the historical stress-temperature dataset.

[0008] Perform a grouping operation on the historical stress-temperature dataset to generate multiple grouped monitoring datasets;

[0009] Wavelength offset feature extraction is performed on the multiple grouped monitoring datasets to obtain multiple core feature values ​​and multiple feature influence ranges;

[0010] Multiple analysis units are established using the influence range of the multiple features, and the multiple analysis units are applied to process the real-time captured optical signal sequence to generate multiple analytical feature sets;

[0011] An integration operation is performed on the multiple analytical feature sets to obtain integrated feature results, and stress measurement values ​​and temperature measurement values ​​are derived based on the integrated feature results.

[0012] Preferably, the acquisition of the identification information of the target fiber Bragg grating sensor includes: receiving external input or automatically identifying the type identifier and installation location information of the target fiber Bragg grating sensor, and transmitting the type identifier and installation location information to the historical storage unit.

[0013] Preferably, the grouping operation on the historical stress-temperature dataset includes: calculating the difference between each data record in the historical stress-temperature dataset, aggregating similar data records into multiple grouped monitoring datasets based on a preset difference threshold, and performing wavelength shift feature extraction on the multiple grouped monitoring datasets.

[0014] Preferably, the wavelength shift feature extraction operation performed on the multiple grouped monitoring datasets includes: parsing the optical wavelength change data of each grouped monitoring dataset, calculating the central trend index of the optical wavelength change data as the feature core value, evaluating the fluctuation range of the optical wavelength change data as the feature influence range, analyzing the feature core value and the feature influence range, and generating multiple parsed feature sets based on the real-time captured optical signal sequence.

[0015] Preferably, establishing multiple analysis units using the multiple feature influence ranges includes: converting each feature influence range into configuration parameters for the analysis unit, initializing the processing capabilities of the multiple analysis units, wherein the configuration parameters are used to adjust the sensitivity settings of the analysis units.

[0016] Preferably, the application of the multiple analysis units to process the real-time captured optical signal sequence includes: acquiring the real-time optical signal sequence of the target fiber Bragg grating sensor, dividing the real-time optical signal sequence into multiple time segments, analyzing the spectral characteristics of each time segment using multiple analysis units, generating multiple analytical feature sets, and integrating them.

[0017] Preferably, the integration operation on the plurality of analytical feature sets includes: calculating the correlation matrix between the plurality of analytical feature sets, merging the plurality of analytical feature sets by weight based on the correlation matrix, and outputting the integrated feature result for deriving stress measurement values ​​and temperature measurement values.

[0018] Preferably, deriving stress and temperature measurements based on the integrated feature results includes: mapping the integrated feature results into stress and temperature components using a physical model, and calculating the final stress and temperature measurements in conjunction with environmental calibration parameters.

[0019] Preferably, the present invention further includes a stress monitoring and temperature measurement system based on fiber Bragg gratings, used to implement the above-described stress monitoring and temperature measurement method based on fiber Bragg gratings. The system includes: an information acquisition module for acquiring identification information of a target fiber Bragg grating sensor; a data query module for performing a query operation in a historical storage unit based on the identification information to determine a historical stress-temperature dataset; a grouping processing module for performing a grouping operation on the historical stress-temperature dataset to generate multiple grouped monitoring datasets; a feature extraction module for performing wavelength offset feature extraction on the multiple grouped monitoring datasets to obtain multiple feature core values ​​and multiple feature influence ranges; an analysis unit construction module for establishing multiple analysis units using the multiple feature influence ranges; a real-time processing module for applying the multiple analysis units to process real-time captured optical signal sequences to generate multiple analytical feature sets; a feature integration module for performing an integration operation on the multiple analytical feature sets to obtain integrated feature results; and a result export module for exporting stress measurement values ​​and temperature measurement values ​​based on the integrated feature results.

[0020] Preferably, the grouping processing module includes a difference calculation submodule and an aggregation submodule. The difference calculation submodule is used to calculate the difference between each data record in the historical stress temperature dataset. The aggregation submodule is used to aggregate similar data records into multiple grouped monitoring datasets based on a preset difference threshold, and transmit the multiple grouped monitoring datasets to the feature extraction module.

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

[0022] This method, through its systematic step design, provides a more comprehensive technical path for stress monitoring and temperature measurement based on fiber Bragg gratings. At the data utilization level, acquiring the identification information of the target fiber Bragg grating sensor and querying historical stress-temperature datasets allows for full utilization of the sensor's accumulated monitoring data. This historical data contains response characteristics under different operating conditions, providing rich reference for subsequent analysis. In-depth mining of historical data makes the analysis process more closely aligned with the sensor's actual operating characteristics, avoiding analytical bias caused by a lack of historical references.

[0023] By grouping the historical stress-temperature dataset into multiple monitoring datasets, data under different operating conditions and environmental environments can be effectively distinguished. Different groups correspond to the sensor's response patterns in specific scenarios, making the grouped data more targeted and facilitating subsequent feature extraction by focusing on the typical features of each group. This grouping method reduces interference from mixed data from different operating conditions, allowing the extracted features to more accurately reflect the stress and temperature effects in the corresponding scenarios, laying a more precise foundation for subsequent real-time signal analysis.

[0024] Wavelength shift feature extraction was performed on multiple grouped monitoring datasets to obtain core feature values ​​and influence ranges, further refining the intrinsic characteristics of the data. Core feature values ​​capture the main trends in wavelength shift within each group, while influence ranges define the normal fluctuation intervals within these trends. This refined feature extraction method provides a more comprehensive representation of historical data features. Compared to the vague feature descriptions in traditional methods, the combination of core values ​​and influence ranges more clearly characterizes the response characteristics of fiber Bragg gratings under different operating conditions, providing a clear reference standard for real-time signal analysis.

[0025] Multiple analysis units are established using feature influence ranges. These units are then applied to process real-time captured optical signal sequences, making real-time signal analysis more targeted. Each analysis unit corresponds to a specific group of feature ranges, accurately matching the operating environment of the real-time signal and avoiding the insufficient adaptability of traditional single analysis models to complex conditions. Through the synergistic effect of multiple analysis units, real-time signals can be analyzed in multiple dimensions and scenarios, resulting in a more realistic feature set that improves the accuracy of real-time signal processing.

[0026] By performing an integration operation on multiple analytical feature sets to obtain integrated feature results, stress and temperature measurements are derived from these results, achieving effective fusion of multi-source information. The integration process synthesizes the analytical results of each analysis unit, balances the differences between different feature sets, and reduces the bias that may arise from single-feature analysis. This integration method fully utilizes the complementarity of historical data features and real-time signal features, enabling the final derived stress and temperature measurements to more accurately reflect actual physical quantity changes, thus enhancing the reliability and stability of the measurement results. Attached Figure Description

[0027] Figure 1 This is a timing diagram of the stress monitoring and temperature measurement method based on fiber Bragg grating described in this invention;

[0028] Figure 2 A flowchart for grouping operations on historical stress-temperature datasets;

[0029] Figure 3 A flowchart for extracting wavelength offset features from grouped monitoring datasets;

[0030] Figure 4 A flowchart for the analysis unit to process real-time captured optical signal sequences. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figure 1 This invention provides a stress monitoring and temperature measurement method and system based on fiber Bragg gratings, the method comprising:

[0033] Accurate measurement of environmental parameters is achieved through a multi-level data processing architecture. The method first obtains the unique identification information of the target fiber Bragg grating sensor, including the device type code and spatial location coordinates. The system accesses a distributed database via a data bus to retrieve a historical monitoring data set associated with this identifier. This set contains wavelength sampling records and corresponding environmental parameters for at least a six-month period. The data processing engine employs a sliding window mechanism to divide the historical dataset along a time dimension, forming continuous data grouping units.

[0034] The feature extraction layer employs digital signal processing techniques to analyze the wavelength distribution characteristics of each data group and calculates the statistical characteristics of the Bragg wavelength shift for each group. For each data group, the system records the arithmetic mean of its wavelength shift as the core feature value, and simultaneously calculates three times its standard deviation as the boundary of the feature influence range. Based on the numerical distribution of the feature influence range, the system dynamically configures the number and parameters of parallel processing channels, with each processing channel corresponding to an independent data analysis unit.

[0035] During the real-time monitoring phase, a high-speed data acquisition card captures the grating reflectance spectral signal at a sampling frequency of no less than 1 kHz, and an analog-to-digital converter converts the analog signal into a digital sequence. Each data analysis unit synchronously processes the segmented signal fragments, extracts spectral features using Fast Fourier Transform, and generates analytical results including parameters such as wavelength offset and spectral width. The data fusion processor receives the output from all channels and synthesizes the final feature vector using a weighted averaging algorithm. The physical conversion module calculates the separated environmental parameter values ​​based on a pre-calibrated stress-temperature coupling matrix.

[0036] Example 1: The processing of sensor identification information begins at the physical interface layer, such as... Figure 2 As shown, the process of grouping historical stress-temperature datasets is as follows: calculate the difference between each data record in the historical stress-temperature dataset; aggregate similar data records into multiple grouped monitoring datasets based on a preset difference threshold; and perform wavelength shift feature extraction on multiple grouped monitoring datasets.

[0037] Specifically, the RS485 serial communication interface establishes a connection between the sensor network and the control host. The transmission protocol includes two key data segments: the device ID segment is a 16-bit binary code containing the device model, production batch, and serial number; the location code segment uses a 32-bit structure, with the first 16 bits recording the installation area number and the last 16 bits storing the three-dimensional spatial coordinates. A CRC-8 checksum is appended to the communication data packet, and the controller performs integrity verification upon receipt. If three consecutive verifications fail, the system automatically sends a re-acquisition command.

[0038] The historical database employs a distributed architecture, stored on a RAID5 protected solid-state drive array. The query component constructs a three-level index structure: the device ID serves as the primary index key, the timestamp forms a B+ tree secondary index, and the installation location generates a geospatial index. For the input target identifier, the retrieval algorithm prioritizes matching the device ID primary key, returning all historical datasets for the corresponding device within milliseconds. The dataset uses a relational data table structure, containing fields such as record number, acquisition time, original center wavelength value, wavelength drift, temperature reference value, and humidity reference value. Each query activates an automatic archiving mechanism, migrating historical records exceeding the set retention period to the archive storage area.

[0039] After loading the query result set, the difference calculation engine performs data normalization. In the preprocessing stage, abnormal null values ​​are detected and filled using linear interpolation, and the original wavelength values ​​are uniformly converted in units. The core algorithm establishes a multi-dimensional feature space: the time interval between adjacent data records, the wavelength difference, and the change in ambient temperature are used as feature vectors. An improved Euclidean distance formula is used to calculate the difference between records, where the weight of the wavelength difference is set to twice the weight of the change in ambient temperature. The system maintains a dynamic threshold controller, which acquires meteorological station data in real time through monitoring ports. When the external temperature change rate exceeds 0.5 degrees Celsius per minute, the threshold recalculation process is automatically triggered. The threshold table is stored in dual-port RAM, and the processor responds to threshold update requests via interrupts.

[0040] Data aggregation processing employs a sliding window mechanism. During initialization, the default window size is configured to 60 consecutive record units, and stack memory is dynamically allocated as a window buffer. The distance calculator sequentially scans the historical dataset, moving one record position at a time. The aggregator activates when the maximum difference between all records within the detection window is lower than the current threshold. The aggregation operation performs data compression, extracting feature values ​​such as the start and end points of time, mean center wavelength, and temperature variation range of the window. Each grouped dataset generates a 128-bit data structure, recording the number of original records and the maximum internal difference value. The grouped output queue uses a priority scheduling strategy, automatically triggering grouped data transmission when three grouped datasets are generated consecutively.

[0041] Error handling mechanisms are implemented throughout the entire process. A timeout interrupt service routine is set up at the communication protocol layer; an error flag is triggered if a complete data packet is not received within 300ms. The database retrieval phase is equipped with a data integrity verification module, which detects storage errors by comparing the consistency of the primary and secondary indexes. The difference calculation core is equipped with a numerical stability control unit, employing a limiting filter to handle drastically fluctuating data. The aggregation processor includes an abnormal group detection algorithm; when the number of records in a single group exceeds twice the normal value, a data segmentation and reconstruction process is initiated.

[0042] The device ID parsing module decomposes the 16-bit code into three functional segments. The high 4 bits map to the device type lookup table, the middle 5 bits identify the manufacturer code, and the low 7 bits correspond to the serial number. Location code parsing uses a spatial mapping algorithm: the first 16 bits are converted into geographic description text through a region mapping table, and the last 16 bits are converted into engineering positioning coordinates based on a preset coordinate system benchmark. All parsing results are temporarily stored in a double-buffered register set, awaiting database call instructions.

[0043] The historical data retrieval process includes a data preprocessing stage. The controller queries the temperature compensation parameter table and loads compensation coefficients based on the current environmental conditions. The raw wavelength values ​​are input into the compensation calculation unit to perform temperature drift compensation and optical power attenuation compensation. Data normalization processing includes sampling time alignment; when the intervals between adjacent records are uneven, a cubic spline interpolation algorithm is used to generate equally spaced data sequences. The preprocessed data is stored in the data preparation area, and timestamps and process status indicators are added.

[0044] The core of the difference calculation includes a hardware acceleration module. A dedicated processor with an arithmetic logic unit handles matrix operations, organizing the feature vectors of data record pairs into a 4×1 matrix. The computational unit pipeline includes: a vector subtraction module, a square calculation module, a weighted accumulation module, and a square root operation module. The result output port connects to a threshold comparator, comparing the calculated result with the dynamic threshold in real time. A flag bit of the comparison result controls the data record flow, guiding data that meets the aggregation conditions into the grouping buffer.

[0045] The group management mechanism employs a double-buffered structure. The group generation process includes: new group initialization, record addition, feature value update, and group closure. The feature value update operation uses an incremental calculation algorithm to avoid repeatedly scanning already processed data. The metadata area of ​​each group data structure records detailed state changes experienced during group formation, including the starting trigger record number, ending record number, group formation duration, and internal record difference statistics, among other historical operation information. After the encapsulated group data is added with a checksum, it is stored in the output queue, and a transmission request signal is sent simultaneously.

[0046] Example 2: Wavelength shift feature extraction is performed on a dedicated signal processing hardware platform. For example... Figure 3 As shown, the process for extracting wavelength shift features from grouped monitoring datasets is as follows: perform wavelength shift feature extraction on multiple grouped monitoring datasets; analyze the optical wavelength change data of each grouped monitoring dataset; calculate the central trend index of the optical wavelength change data and use it as the core feature value; evaluate the fluctuation range of the optical wavelength change data and use it as the feature influence range; analyze the core feature value and the feature influence range to generate multiple analytical feature sets from the real-time captured optical signal sequence.

[0047] Specifically, the platform includes multi-channel data input interfaces, each connected to the storage buffer of the grouped monitoring dataset. The data loading controller manages the data transmission process, using DMA (Direct Memory Access) to directly move the grouped data to the local memory of the processing core. The preprocessing unit first performs signal conditioning, baseline calibration of the input wavelength sequence. The calibration process references the device's factory calibration parameters, reading the zero-point offset from non-volatile memory. The conditioned data then enters the digital filtering stage, configuring a programmable FIR (Finite Impulse Response) filter bank and automatically selecting filter coefficients based on the temporal characteristics of the data groups.

[0048] The feature core value calculation module uses a median filtering algorithm to process the wavelength sequence. The filter window size is dynamically adjusted, with an initial default value of 11 data points, and the window size changes linearly with the time span of data grouping. The filter processor includes a data sorting array and selection logic, and the median value of the output sequence is used as the preliminary core value. The post-processing unit performs a smoothing operation on the preliminary core value, using a moving average algorithm to calculate the arithmetic mean of five consecutive preliminary core values. The final feature core value is stored in a dual-port register, along with a timestamp and group identifier.

[0049] The fluctuation range analysis module receives the same filtered wavelength sequence. This module establishes a statistical distribution model of the data points, first constructing a histogram of wavelength offsets. The histogram unit contains 64 counting channels, with the channel width dynamically set based on the maximum and minimum values ​​of the data groups. The statistical processor scans the histogram distribution to locate the minimum coverage interval containing 95% of the data points. The interval boundary detection algorithm employs a bidirectional scanning mechanism, expanding outwards from the distribution center until the cumulative count reaches a threshold. The interval width value serves as the fundamental measure of the feature's influence range.

[0050] The gradient analysis unit processes wavelength sequence variation characteristics in parallel. The differential calculator uses the central difference method to calculate the first derivative of each data point. The mutation detection logic monitors the absolute value of the derivative sequence; when the derivative values ​​of three consecutive points exceed a set threshold, a waveform anomaly flag is triggered. The response mechanism automatically expands the analysis window, incorporating 20% ​​more sampling points before and after the current data point into the recalculation. The recalculation process is executed iteratively until the waveform stability condition is met.

[0051] The feature output interface is designed with a bidirectional communication structure. The output buffer adopts a ping-pong buffer architecture, with two 512-bit buffer blocks working alternately. When data is written to the main buffer, the backup buffer transmits it outward through a high-speed serial interface. The transmitted data packet contains a 32-bit floating-point number representing the feature core value, a 32-bit floating-point number representing the feature's influence range, a data packet identifier, and status flags. The status flags include data validity indicators, abnormal waveform markers, and calculation process check codes.

[0052] The hardware platform includes a temperature compensation subsystem. A temperature sensor monitors the chip junction temperature in real time, and the ADC (Analog-to-Digital Converter) acquires temperature data at a rate of 100 times per second. The compensation calculation unit stores the temperature-accuracy correction curve, and automatically activates the compensation program when a temperature change exceeding 2 degrees Celsius is detected. The compensation value is applied to the calculated results of the feature core value and the area of ​​influence; the correction amount is obtained through a lookup table method.

[0053] An error detection mechanism is implemented throughout the entire feature extraction process. The input data verification unit verifies the integrity of the grouped data, checking the reasonableness of the number of records and their numerical ranges. The calculation process monitor employs a dual-core redundant design: the main processor performs the actual calculations, while the coprocessor verifies the results. When the difference between the outputs of the two processors exceeds the allowable error, the system automatically initiates a recalculation process. The output interface includes a cyclic redundancy check (CRC) generator, appending a 16-bit CRC to each data packet.

[0054] The baseline calibration in the preprocessing stage employs a multi-level compensation strategy. Level 1 calibration uses the device's factory parameters, Level 2 calibration references historical data trends, and Level 3 calibration incorporates environmental temperature correction. Calibration parameters are stored in ferroelectric memory, supporting millions of erase / write cycles. The calibration process includes an automatic zeroing procedure; when fluctuations at ten consecutive data points are detected to be less than a threshold, automatic zero-point update is triggered.

[0055] The hardware implementation of the median filtering algorithm employs a parallel sorting network. The data input register is connected to a comparator array, and sorting is completed through multi-stage comparison and exchange operations. Selection logic extracts the center position value of the sorted sequence, supporting direct selection of odd-numbered data points and averaging of even-numbered data points. The filter window size controller is connected to the system clock, adjusting the time range covered by the window according to the data sampling interval.

[0056] The statistical distribution analysis module includes a histogram update mechanism. Each time a new data point is input, the histogram counter performs an incremental update. Interval boundary detection uses a binary search algorithm, searching from the distribution center outwards for boundary points that meet the cumulative probability requirements. The search process records the boundary movement trajectory; when new data causes a boundary change exceeding 5%, a complete recalculation process is triggered.

[0057] The waveform anomaly handling adopts an event-driven architecture. The mutation detector outputs an interrupt signal, and the processor suspends the current computation task after responding to the interrupt. The context saving unit stores the current computation state, and the anomaly handler loads the extended data window. After recalculation, the original context is restored, and subsequent processing flows continue. The event log records the time, location, and processing result of each anomaly trigger.

[0058] The output interface's ping-pong buffer management employs a state machine control. Write and read pointers operate independently, and an address comparator monitors the buffer status. When the main buffer is full or a forced transfer command is received, the switching control logic automatically swaps the buffer roles. The transmission protocol includes a start-of-packet flag, length field, payload, and end-of-packet flag, conforming to industry-standard serial communication specifications.

[0059] The temperature compensation subsystem establishes a three-dimensional lookup table. Compensation parameters are organized according to three dimensions: chip temperature, ambient humidity, and operating voltage. The parameter interpolation unit uses a bilinear interpolation algorithm to calculate the actual compensation value. The compensation application module includes multipliers and adders to perform scaling transformations and zero-point shifting on the feature calculation results. The compensation value update cycle is synchronized with the temperature sampling rate to ensure real-time compensation effectiveness.

[0060] The computational verification mechanism employs a difference-tolerance strategy. The main and coprocessors run the same algorithm independently, with a programmable error range set for the result comparator. When the difference is within the allowable range, the main processor's result is used; when the difference exceeds the limit, an arbitration procedure is initiated. The arbitrator uses historical data trends for auxiliary judgment, selecting the computational result that best matches the trend. The verification process is recorded in the system log, containing detailed data for each comparison.

[0061] Example 3: The process of establishing the analysis unit begins with the conversion mechanism of configuration parameters. For example... Figure 4 As shown, the process of the analysis unit in processing the real-time captured optical signal sequence is as follows: acquiring the real-time optical signal sequence of the target fiber Bragg grating sensor; dividing the real-time optical signal sequence into multiple time segments; analyzing the spectral characteristics of each time segment using multiple analysis units; generating multiple analytical feature sets and integrating them.

[0062] Specifically, the system receives feature influence range data through a parameter conversion interface, which is stored in a shared register set as a 32-bit floating-point number. The conversion process uses mathematical relationships to calculate configuration parameters, expressed by the following formula:

[0063]

[0064] in, These are configuration parameters, representing the sensitivity adjustment value of the analysis unit, expressed as a dimensionless proportionality coefficient. It is the basic scaling constant, taken from the preset system configuration table, and its value range is fixed from 0.8 to 1.2; It is the attenuation coefficient, which is dynamically adjusted according to the device type and stored in non-volatile memory; This is the input value representing the range of influence of the features, reflecting the wavelength fluctuation of the data group. The parameter conversion processor executes this formula to calculate the result, which is quantized to 8-bit integer precision and used as the core configuration value. The conversion process includes a numerical verification step to check whether the input range is within the valid interval; if it exceeds the threshold, a default value loading mechanism is triggered.

[0065] Initializing the analysis unit involves resource allocation and function settings. The system controller dynamically allocates processing resources based on configuration parameter values. Each analysis unit is initialized independently, including the following operations: allocating a dedicated memory buffer for the unit, the buffer size being proportional to the configuration parameter value, with a minimum allocation of 1KB and a maximum expansion to 128KB; constructing a Hamming window function structure, the window function width being inversely proportional to the configuration parameter, for example, when the configuration parameter increases, the window width automatically shrinks to half of its original value; loading the digital filter coefficient library, selecting a bandpass filter with matching parameter values ​​from pre-stored filtering schemes; and finally, starting a self-test program to verify memory integrity and filter initialization status. All units form a parallel processing array, interconnected via a cross-switch matrix. After initialization, the unit enters a ready state, awaiting real-time signal processing trigger signals.

[0066] The real-time signal processing stage is initiated via a high-speed data acquisition module. The optical signal sequence from the target fiber Bragg grating sensor is captured by a photoelectric converter with a sampling frequency set to 2kHz. An ADC converter quantizes the analog signal into a 16-bit digital sequence. This sequence is input to a signal segmentation controller, where the segmentation algorithm operates based on a time window mechanism: the continuous signal is divided into fixed-length segments in 50-millisecond increments, each segment containing 100 uniformly sampled points. A time synchronizer adds timestamps to ensure each segment is associated with precise time coordinates. The segmented sequence is distributed to an analysis unit array, with unit allocation employing a round-robin scheduling strategy, allowing each unit to process one time segment independently.

[0067] The internal processing flow of the analysis unit comprises multiple functional modules. The input signal segment first passes through a configurable digital filter, the filter type determined by coefficients loaded during initialization, performing frequency-selective filtering. Next, it enters the spectral characteristic analysis core, employing a peak detection algorithm to locate characteristic peaks in the reflectance spectrum: performing local maxima search on the filtered signal to calculate the wavelength coordinates and intensity values ​​corresponding to the peaks; simultaneously analyzing spectral width characteristics, quantizing the signal extension range by calculating the full width at half maximum (FWHM). The processing results for each time segment are encapsulated as an analytical feature set, including wavelength offset values, spectral width values, and confidence indices. The output interface uses a high-speed serial protocol to transmit data, along with a unit identifier and processing timestamp, providing input for subsequent integration operations.

[0068] Error handling and safety mechanisms are integrated throughout the analysis unit's lifecycle. An out-of-range detector is included during the configuration transition phase; when an input within the feature's influence range becomes abnormal, the system switches to a redundant parameter table to load backup configuration parameters. The unit initialization process includes hardware diagnostics: a memory tester performs write / read checks, and a filter logic verifier simulates input / output consistency. Signal integrity monitoring is embedded in real-time processing: an ADC sampling sequence checksum generator detects transmission errors and triggers a resampling command; the peak detection algorithm's tolerance module filters out abnormal noise points to prevent misjudgments. All error events are logged to the system log, with the log structure including a timecode, error type, and recovery action identifier.

[0069] The performance optimization of the analysis unit is based on dynamic parameter adjustment. The output of the configuration parameter converter is connected to a parameter feedback loop. When the characteristics of the real-time signal drift, the system automatically adjusts the values ​​of constants a and b by retrieving the optimal parameter combination from a historical performance database. The initialization manager monitors memory usage efficiency and uses a defragmentation algorithm to optimize buffer allocation. The real-time processing module includes a throughput balancing mechanism; when the segment processing delay exceeds 50 microseconds, the number of allocated units is reduced or the time segment length is compressed. All optimization processes are executed in the background, without affecting the continuity of front-end data processing.

[0070] The implementation details of the signal segmentation algorithm include boundary handling strategies. For signal sequences that are not integer multiples of their length, the tail segments are padded with zeros. An overlapping window mechanism handles the transition between consecutive segments: when the intervals between adjacent segments are uneven, the controller automatically applies a 10% overlap region to reduce boundary effects; the signal data in the overlapping part is recalculated but only output once. The segmentation accuracy calibration unit is connected to the system clock source to compensate for time drift errors in real time, ensuring that the time length deviation of each segment is less than 1 microsecond.

[0071] The feature set generation process includes quality control. Each analysis unit's output includes a confidence calculation module: based on signal-to-noise ratio and peak intensity data, it assesses the reliability of the results and outputs a scale value between 0 and 1. The feature set structure uses a fixed 16-byte data packet: the first 4 bytes store the wavelength offset, the middle 4 bytes are the spectral width value, the following 4 bytes contain the time and unit identifier, and the last 4 bytes record the confidence score and checksum. The transmission protocol implements data integrity protection: a 16-bit CRC checksum is added, and the receiving end requests retransmission if verification fails.

[0072] The collaborative operation of the unit array is coordinated by a central scheduler. The scheduler manages the unit resource pool, and when a new signal sequence arrives, it queries the unit's readiness status and allocates segments to be processed. The load balancing module analyzes the processing time history of each unit and dynamically adjusts the allocation weights to ensure even task distribution. Processing results are uniformly aggregated into an output buffer queue, configured to hold at least 100 parsed feature sets to prevent data loss.

[0073] Temperature and environmental factor compensation mechanisms are integrated into the analysis workflow. Temperature sensors monitor the internal temperature of the equipment in real time, and the compensation calculator corrects filter parameters and configuration parameter values ​​based on temperature changes. Compensation data is stored in a multidimensional lookup table, categorized by degree Celsius intervals to ensure calculation accuracy. When environmental changes exceed a set threshold, a re-initialization operation is automatically triggered to maintain system consistency.

[0074] Implemented on an embedded processing platform, the hardware platform includes a multi-core processor and an FPGA (Field-Programmable Gate Array) accelerator. The FPGA handles signal segmentation and peak detection algorithms, while the multi-core CPU executes unit initialization and configuration conversion logic. The platform resource manager optimizes task scheduling, supporting millisecond-level response times for the real-time operating system. All components are interconnected via a system bus, and data flow is pipelined to maximize throughput efficiency.

[0075] Example 4: The feature integration processor receives a sequence of parsed feature sets from multiple analysis units. The input interface employs a multiplexed architecture, supporting simultaneous reception of data streams from 32 independent channels. Each parsed feature set contains a data packet with a fixed format: timestamp, wavelength offset, spectral width value, confidence score, and unit identifier. The data packet is stored in the input buffer after CRC verification. The buffer is organized as a circular queue with a capacity designed to store 200 feature sets. When the buffer reaches 75% occupancy, the integration processing flow is triggered.

[0076] The correlation matrix construction engine first performs data alignment. The time synchronization module scans the timestamps of all feature sets and establishes a timeline based on the earliest timestamp. The interpolation processor performs linear interpolation on the asynchronous data points to generate a time-aligned feature vector sequence. The matrix generator creates an N×N symmetric matrix structure, where N represents the number of feature sets in the current processing cycle. Matrix element calculation uses a sliding window mechanism: for each pair of feature sets (i,j), the correlation coefficient of their wavelength offset sequence at 20 consecutive time points is calculated. The correlation coefficient calculation process includes mean centering and normalization, and the output value is limited to the range of -1 to +1.

[0077] The weight calculation module performs matrix feature analysis. This module uses the Jacobi iterative algorithm to calculate the eigenvalues ​​and eigenvectors of the matrix. The main eigenvector extractor locates the eigenvector component corresponding to the largest eigenvalue. The weight allocator normalizes the eigenvector components into weight coefficients, ensuring that the sum of all weights equals 1. The weight coefficients are temporarily stored in a register set, with each coefficient associated with a corresponding feature set identifier.

[0078] The weighted merging operation is performed in the data fusion core. This core contains an array of parallel multipliers and accumulators. For each feature dimension (wavelength offset and spectral width), weighted calculations are performed independently: feature values ​​are read from the input buffer, multiplied by the corresponding weight coefficients, and the result is fed into the accumulator. The accumulation process continues until all feature sets have been processed. The final output is the integrated feature result, containing two main components: the integrated wavelength offset and the integrated spectral width value. Each component is accompanied by a quality metric reflecting the uniformity of the weight distribution.

[0079] The anomaly detection unit monitors the integration process in real time. The residual calculator compares the contribution value of each feature set with the difference in the integration result. The difference threshold setter dynamically adjusts the threshold value based on historical residual data. When the residual of a feature set exceeds three times the standard deviation three times consecutively, a reweighting procedure is triggered. The reweighting controller temporarily reduces the weight coefficient of that feature set by 50% of its original value, while proportionally increasing the weights of other feature sets. The adjusted weights are then re-entered into the weighted merging process.

[0080] The output interface generates a standardized integrated feature result structure. This structure contains the following fields: integrated wavelength offset (32-bit floating-point number), integrated spectral width (32-bit floating-point number), time range start marker, time range end marker, number of feature sets involved, and weight distribution entropy value. It also outputs auxiliary data: the original contribution value of each feature set, the adjusted weight coefficients, and a residual size marker. All data is encapsulated into a 128-bit data packet with an appended 32-bit error check code.

[0081] A quality control mechanism is implemented throughout the entire process. The input verification unit verifies the continuity of the timestamps of the feature sets. When a time jump exceeds twice the sampling interval, a virtual feature set is inserted to fill the gap. The matrix symmetry checker verifies the mathematical properties of the correlation matrix and automatically corrects asymmetric elements. The weight rationality monitor detects the distribution of weight coefficients and triggers weight smoothing when the maximum weight exceeds 0.5. The output verification module adopts a dual-computation channel redundancy design to compare the differences between the results of the two independent channels. Assuming that the current processing cycle receives five analytical feature sets (F1 to F5), their wavelength offset data, after time alignment, forms the following sequence, as shown in Table 1.

[0082] Table 1 shows the wavelength offset data, which, after time alignment, forms the following sequence (unit: pm).

[0083]

[0084] The correlation matrix calculation module analyzes the data from these 20 time points and generates a 5×5 symmetric matrix. Example matrix data is as follows: the correlation coefficient between F1 and F3 is 0.92, the correlation coefficient between F1 and F4 is 0.35, and the correlation coefficient between F2 and F5 is 0.88. Feature analysis reveals that the largest eigenvalue is 3.45, corresponding to the eigenvector [0.28, 0.25, 0.29, 0.08, 0.10]. The normalized weight coefficients are assigned as follows: F1: 0.28, F2: 0.25, F3: 0.29, F4: 0.08, F5: 0.10.

[0085] The weighted merging process calculates the integrated wavelength offset at time t20 as: 155.1×0.28 + 154.3×0.25 + 154.8×0.29 + 157.2×0.08 + 154.0×0.10 = 154.82 pm. Residual detection reveals a significant difference between the contribution value of the F4 feature set (157.2 pm) and the integrated result (154.82 pm), exceeding the current threshold. The system initiates a reweighting procedure, reducing the F4 weight to 0.04 and proportionally increasing the weights of other feature sets. After recalculation, the integrated result is 154.79 pm, and the residual is reduced to within the acceptable range.

[0086] The system maintains a historical weight database, recording the weight allocation scheme for each processing cycle. When the same feature set is downweighted for ten consecutive cycles, a feature set health status assessment process is triggered. The assessment results are fed back to the front-end analysis unit controller to guide the parameter adjustment of the analysis unit.

[0087] The time management unit coordinates the entire processing sequence. Each processing cycle is set to a fixed 100-millisecond window and includes four stages: data reception, matrix construction, weight calculation, and weighted merging. Hardware timers enforce timeout control; when a stage times out, it automatically switches to simplified processing mode. Simplified mode uses a preset fixed weight scheme to ensure real-time requirements.

[0088] Output data packets are sent via a dual-channel transmission mechanism. The main channel uses a high-speed serial interface to transmit complete data packets; the auxiliary channel generates a simplified version, containing only the integrated wavelength offset and time stamp. The receiving end selects the data channel according to application requirements, and control commands are transmitted to the integrated processor via a feedback link.

[0089] The environmental adaptability design includes a temperature compensation module. When the processor's internal temperature changes by more than 5 degrees Celsius, the compensation procedure is activated: it adjusts the normalization parameter in the correlation coefficient calculation to compensate for calculation errors caused by temperature. The compensation parameters are stored in a temperature-parameter lookup table, with a set of calibration values ​​stored at each degree Celsius interval.

[0090] Example 5: The physical conversion module receives the integrated feature results from the feature integration stage. This result structure contains the integrated wavelength offset and integrated spectral width values ​​within the time window. The conversion operation begins with the conversion matrix loading process: the system controller reads the basic conversion matrix from the device feature parameter memory. This matrix is ​​a 4×4 constant coefficient matrix stored in the parameter area using flash memory. Parity checking is performed during matrix loading; if the check fails, a redundant matrix from the backup area is automatically loaded. The conversion processor initializes the vector space, organizing the integrated wavelength offset, integrated spectral width, ambient temperature reference value, and atmospheric pressure reference value into input feature vectors. The matrix multiplier performs multiplication of the feature vectors and the conversion matrix, outputting a transition vector containing two elements: initial stress component and initial temperature component. The computation process is completed using a fixed-point arithmetic accelerator, ensuring millisecond-level response speed.

[0091] The environmental compensation subsystem operates in three levels. Primary compensation uses equipment characteristic parameters to retrieve the correction curve that best matches the current integrated characteristic result from the calibration database. This curve is generated based on calibration data from a temperature-controlled environment at the time of equipment delivery and is stored in a lookup table containing 256 discrete calibration points. Secondary compensation incorporates real-time environmental monitoring data: a high-precision atmospheric pressure sensor samples air pressure values ​​at a frequency of 10Hz, and a digital temperature sensor monitors the equipment casing temperature. The compensation calculation engine calculates dynamic compensation amounts based on the temperature gradient and the rate of change of air pressure. The temperature gradient is taken as the slope of change over the past 60 seconds, and the rate of change of air pressure is taken as the average of the most recent 30 sampling points. Tertiary compensation activates the aging correction mechanism: a counter records the cumulative working time during equipment operation, and the next stage of compensation coefficients is automatically loaded every 1000 hours. The compensation amount is applied to each component of the transition vector, using an additive correction mode to superimpose it onto the initial component value.

[0092] The calibration parameter management system maintains a three-tiered storage structure. The static parameter area stores the underlying parameters that remain unchanged throughout the lifespan, including sensor material coefficients and grating constants. The dynamic parameter area stores field calibration values ​​and accepts correction values ​​input by the user via the human-machine interface. The temporary parameter area caches compensation coefficients generated through self-learning. Parameter retrieval follows a priority rule: when the system timestamp is less than 30 days after the latest calibration, dynamic parameters are used first; otherwise, static parameters are used. Parameter updates employ a double-buffering strategy: new parameter groups undergo 72 hours of stability testing in the background verification area before being switched to the working area to take effect. All parameter access is logged in detail, including retrieval time, operator identifier, and version checksum.

[0093] A multi-stage filtering mechanism is designed for the output processing stage. A timing controller triggers output sequence generation with a fixed 100-millisecond cycle, inputting the latest compensated stress and temperature component values ​​in each cycle. A primary sliding filter maintains a data window of 10 consecutive cycles, calculating the median within the window as the initial output. A secondary time-domain filter analyzes historical trends: recording the measurement sequence of the past 30 minutes; when the current value deviates from the trend prediction range, the filter intensity is automatically increased. A third-stage adaptive filter adjusts according to environmental stability: when air pressure fluctuations exceed 1 hPa per hour, the filter window is expanded to 30 cycles; when the temperature change rate is less than 0.1 degrees Celsius per minute, the default 10-cycle window is restored. The final output value undergoes amplitude limiting processing, restricting the output change rate to no more than 10% of the previous value to eliminate abrupt interference.

[0094] The numerical output interface implements a dual-mode transmission channel. The analog output circuit adopts a 4-20mA current loop design. The current converter receives a 16-bit numerical input from the digital processor and generates a proportional current signal through a programmable current source. The voltage protection unit monitors line abnormalities and automatically switches protection modes when an open circuit or short circuit is detected. The digital interface implements the Modbus RTU (Modbus Remote Terminal Unit) protocol stack, and the physical layer uses RS485 differential signal transmission. The protocol processing unit encapsulates two types of data frames: the standard frame contains the latest stress and temperature values; the extended frame adds time stamps and environmental parameters. The transmission baud rate is adaptively adjusted: it attempts at 115200bps during initialization, and gradually reduces to 9600bps if communication fails. The interface address is configured by a DIP switch, supporting device address settings from 1 to 247.

[0095] The detailed operation of the compensation calculation engine involves multiple parallel processing branches. The temperature compensation thread continuously receives ambient temperature samples and calculates a moving average every 5 seconds. The compensation table retrieval uses a binary search algorithm to locate the compensation coefficient corresponding to the current temperature range. The air pressure compensation module implements altitude correction: based on a standard atmospheric model, it converts air pressure values ​​into altitude and corrects stress values ​​using a preset altitude-compensation table. The axial compensation processor specifically handles stress components; when temperature component changes exceed a threshold, it applies a Poisson's ratio correction algorithm to adjust the sensitivity coefficient of the stress value. All compensation calculations adopt a pipelined architecture, with four compensation modules processing different dimensions of correction in parallel, and finally performing vector superposition at the data aggregation node.

[0096] The equipment health monitoring function is integrated into the output end. The anomaly detector analyzes the raw data sequence before filtering and identifies abnormal patterns including: constant values ​​lasting for more than 10 cycles, exceeding the measurement range by 125%, and fluctuations exceeding 50% between adjacent cycles. The monitoring results output diagnostic codes: 0x01 indicates suspected sensor malfunction, 0x02 indicates abnormal environmental compensation, and 0x04 marks a communication interface error. Diagnostic information is output through a separate digital port and simultaneously stored in a non-volatile error log memory. The system periodically triggers a self-test program: every 24 hours, it comprehensively verifies the integrity of the transformation matrix by substituting preset eigenvectors to verify the output results; weekly, it performs zero-point drift calibration of the pressure sensor, achieving automatic calibration through a closed reference chamber.

[0097] Real-time clock management ensures accurate time stamping. A backup battery-powered clock circuit maintains independent timing, synchronizing with the main controller clock every 10 minutes. The time stamp format uses the industry-standard Unix timestamp, including a 32-bit second counter and a 32-bit microsecond counter. The time information in the output data packet is associated with the GPS receiver: when the GPS signal is valid, UTC standard time is used; after signal loss, it automatically switches to crystal oscillator timekeeping mode, maintaining accuracy with a daily error of less than 1 second.

[0098] The analog output channel features a multi-layered safety protection mechanism. A current loop monitor continuously monitors the output current value, triggering an alarm when the actual current deviates from the set value by more than 0.5mA. The overvoltage protection circuit employs a transient voltage suppression diode array to absorb surge energy up to 200 joules. Reverse connection protection allows for incorrect power polarity connection without damaging the device. A hot-swap protection module maintains output stability during interface insertion and removal, preventing value fluctuations.

[0099] The adaptive updating of aging parameters employs a machine learning strategy. A historical database continuously records the deviations between operating parameters and actual measurements. When a deviation trend under a specific temperature and pressure combination persists for 72 hours, the analysis engine generates a new compensation coefficient proposal. After manual review and approval via an interface, the proposal is automatically updated in the dynamic parameter area. The update operation follows a phased deployment principle: it is first tested on non-critical equipment for 240 hours to verify stability before being rolled out to all equipment. All parameter changes form a complete traceability chain, supporting parameter version rollback at any point in time.

[0100] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0101] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for stress monitoring and temperature measurement based on fiber Bragg gratings, characterized in that, The method includes the following steps: The identification information of the target fiber Bragg grating sensor is obtained, and a query operation is performed in the historical storage unit based on the identification information to determine the historical stress-temperature dataset. Perform a grouping operation on the historical stress-temperature dataset to generate multiple grouped monitoring datasets; Wavelength offset feature extraction is performed on the multiple grouped monitoring datasets to obtain multiple core feature values ​​and multiple feature influence ranges; Multiple analysis units are established using the influence range of the multiple features, and the multiple analysis units are applied to process the real-time captured optical signal sequence to generate multiple analytical feature sets; An integration operation is performed on the multiple analytical feature sets to obtain integrated feature results, and stress measurement values ​​and temperature measurement values ​​are derived based on the integrated feature results. The grouping operation performed on the historical stress-temperature dataset includes: Calculate the degree of difference between each data record in the historical stress-temperature dataset, aggregate similar data records into multiple grouped monitoring datasets based on a preset difference threshold, and perform wavelength shift feature extraction on the multiple grouped monitoring datasets; The wavelength shift feature extraction operation performed on the multiple grouped monitoring datasets includes: The light wavelength change data of each group monitoring dataset is analyzed, the central trend index of the light wavelength change data is calculated as the core feature value, the fluctuation range of the light wavelength change data is evaluated as the feature influence range, and the core feature value and feature influence range are analyzed to generate multiple analytical feature sets based on the real-time captured light signal sequence. The application of the multiple analysis units to process the real-time captured optical signal sequence includes: The real-time optical signal sequence of the target fiber Bragg grating sensor is acquired, the real-time optical signal sequence is divided into multiple time segments, the spectral characteristics of each time segment are analyzed by multiple analysis units, multiple analytical feature sets are generated, and they are integrated.

2. The stress monitoring and temperature measurement method based on fiber Bragg grating according to claim 1, characterized in that, The acquisition of the identification information of the target fiber Bragg grating sensor includes: The system receives external input or automatically identifies the type identifier and installation location information of the target fiber Bragg grating sensor, and transmits the type identifier and installation location information to the historical storage unit.

3. The stress monitoring and temperature measurement method based on fiber Bragg grating according to claim 1, characterized in that, The method of establishing multiple analysis units using the influence range of the multiple features includes: The influence range of each feature is converted into configuration parameters for the analysis unit, and the processing capabilities of multiple analysis units are initialized. These configuration parameters are used to adjust the sensitivity settings of the analysis units.

4. The stress monitoring and temperature measurement method based on fiber Bragg grating according to claim 1, characterized in that, The integration operation on the multiple parsed feature sets includes: Calculate the correlation matrix among multiple analytical feature sets, and then weight and merge the multiple analytical feature sets based on the correlation matrix to output the integrated feature result for deriving stress measurement values ​​and temperature measurement values.

5. The stress monitoring and temperature measurement method based on fiber Bragg grating according to claim 4, characterized in that, The process of deriving stress and temperature measurements based on the integrated feature results includes: A physical model is used to map the integrated feature results into stress and temperature components, and the final stress and temperature measurements are calculated by combining environmental calibration parameters.

6. A stress monitoring and temperature measurement system based on fiber Bragg gratings, characterized in that, For implementing the stress monitoring and temperature measurement method based on fiber Bragg grating as described in any one of claims 1 to 5, the system comprises: The information acquisition module is used to acquire the identification information of the target fiber Bragg grating sensor; The data query module is used to perform a query operation in the historical storage unit based on the identification information to determine the historical stress temperature dataset; The grouping processing module is used to perform grouping operations on the historical stress-temperature dataset to generate multiple grouped monitoring datasets; The feature extraction module is used to perform wavelength offset feature extraction operations on the multiple grouped monitoring datasets to obtain multiple core feature values ​​and multiple feature influence ranges; An analysis unit construction module is used to establish multiple analysis units using the influence ranges of the multiple features; The real-time processing module is used to process the real-time captured optical signal sequence using the multiple analysis units to generate multiple analytical feature sets; The feature integration module is used to perform an integration operation on the multiple parsed feature sets to obtain integrated feature results; the result export module is used to export stress measurement values ​​and temperature measurement values ​​based on the integrated feature results.

7. The stress monitoring and temperature measurement system based on fiber Bragg grating according to claim 6, characterized in that, The packet processing module includes: The difference calculation submodule is used to calculate the difference between each data record in the historical stress-temperature dataset; The aggregation submodule is used to aggregate similar data records into multiple grouped monitoring datasets based on a preset difference threshold, and then transmit the multiple grouped monitoring datasets to the feature extraction module.

Images