An abnormal temperature data identification method based on gallium arsenide optical fiber temperature measurement

Abstract

The present application relates to the technical field of optical fiber temperature measurement analysis, and particularly relates to an abnormal temperature data identification method based on gallium arsenide optical fiber temperature measurement, comprising the following steps: synchronously collecting temperature time series and reflection spectrum sequence of a measured point by a gallium arsenide temperature measurement optical fiber sensor, and extracting a band-pass spectrum shape from the reflection spectrum sequence to obtain a pure shape sequence decoupled from temperature physical quantities; inputting the pure shape sequence into a preset shape-degradation mapping framework to output a degradation index; performing bivariate conjugate determination on the temperature time series and the degradation index to mark a type I implicit abnormality, a type II explicit abnormality or a type III coupled abnormality. The present application not only retains the monitoring capability for direct faults such as overheating, but also can mark potential risks in advance when the temperature has not exceeded the limit but the degradation signal has significantly increased, thereby providing more detailed judgment basis for subsequent differentiated alarm level control, redundancy switching strategy and protection logic triggering.

Description

Technical Field

[0001] This invention relates to the field of fiber optic temperature measurement and analysis technology, and in particular to a method for identifying abnormal temperature data based on gallium arsenide fiber optic temperature measurement. Background Technology

[0002] Fiber optic temperature sensors, due to their strong resistance to electromagnetic interference, long-distance deployment capability, and high-resolution temperature measurement ability, have been widely used in heat-sensitive scenarios such as high-voltage equipment, power modules, energy storage units, and aerospace components. In recent years, with the increasing requirements for operational safety, traditional temperature measurement mechanisms that rely solely on temperature threshold judgment have gradually revealed problems such as response lag and missed detection of potential hazards. Especially in the early stages of thermal fatigue and interface aging, the temperature may still be within the normal range, causing potential degradation to go undetected in a timely manner.

[0003] Gallium arsenide (GaAs) materials, due to their excellent wide-band response characteristics and high-temperature stability, are increasingly being applied in high-precision fiber optic temperature measurement systems, making it possible to obtain stable reflection spectra. However, existing technologies mostly focus on peak shifts or changes in absolute intensity of the reflection spectrum, failing to fully explore the degradation characteristics inherent in the spectral shape itself. Furthermore, existing temperature measurement systems generally lack the ability to jointly model spectral morphology and temperature values, resulting in weak identification capabilities for latent degradation phenomena. In addition, regarding anomaly identification, most systems determine anomalies based solely on single-variable temperature data, making it difficult to accurately distinguish between latent degradation, explicit temperature rise, or complex conditions combining both, affecting the accuracy and timeliness of subsequent alarm and protection strategies. Summary of the Invention

[0004] This invention provides an abnormal temperature data identification method based on gallium arsenide fiber optic temperature measurement. This method enables decoupled analysis of temperature and spectral shape, has the ability to extract multi-layer degradation features, and can perform bivariate anomaly classification, thereby improving the detection capability of early degradation and complex anomaly states.

[0005] A method for identifying abnormal temperature data based on gallium arsenide fiber optic temperature measurement includes the following steps: S1, the temperature time series and reflection spectrum sequence of the measured point are synchronously acquired by the gallium arsenide temperature measurement fiber optic sensor, and the bandpass spectrum shape of the reflection spectrum sequence is extracted to obtain a pure shape sequence decoupled from the temperature physical quantity. The pure shape sequence is the first derivative vector of the reflection spectrum at the same time after normalization. S2, input the pure shape sequence into a preset shape-degradation mapping framework and output a degradation index. The shape-degradation mapping framework includes sequentially connected spectral entropy increase operation, thermal fatigue echo template convolution operation and negative step sharpness operation, and the value of the degradation index increases monotonically with the increase of spectral disorder, echo energy and negative step slope. S3. Perform bivariate conjugate determination on the temperature time series and degradation index. When the temperature time series is within the normal temperature range and the degradation index is greater than the degradation threshold, it is marked as a Type I latent anomaly. When the temperature time series exceeds the normal temperature range and the degradation index is not greater than the degradation threshold, it is marked as a Type II explicit anomaly. When the temperature time series exceeds the normal temperature range and the degradation index is greater than the degradation threshold, it is marked as a Type III coupled anomaly. The Type I, Type II, and Type III marking results are used to trigger subsequent differentiated protection actions.

[0006] Optionally, S1 further includes transmitting scanning light to the gallium arsenide temperature sensing fiber optic sensor via a broadband tunable laser and simultaneously receiving its reflected light signal.

[0007] Optionally, the reflected light signal can be decomposed into an intensity demodulation channel and a spectral demodulation channel: The intensity demodulation channel is converted into an electrical signal by a photodetector, and then calculated by a temperature calibration model to obtain continuous temperature time series data; The spectral demodulation channel acquires a time-indexed reflectance spectrum sequence through a spectrometer, with each reflectance spectrum including light intensity values ​​at multiple wavelength sampling points.

[0008] Optionally, bandpass spectral shape extraction is performed on each reflection spectrum in the reflection spectrum sequence: Select a reference spectral bandpass window that is adjacent to the temperature-sensitive peak of gallium arsenide material and is insensitive to temperature changes; The reflectance spectrum data within the reference spectral bandpass window is extracted and subjected to minimum-maximum normalization to eliminate the influence of fluctuations in the absolute value of light intensity. Calculate the first derivative of the minimum-maximum normalized spectrum to form a vector describing the local rate of change of the spectrum. This vector is the pure shape sequence decoupled from the temperature physical quantity.

[0009] Optionally, the spectral entropy increase operation specifically includes: calculating the information entropy of the first derivative vector as the initial disorder value; comparing the initial disorder value at the current moment with a benchmark entropy value obtained from historical normal data statistics, and calculating its relative increment, wherein the relative increment serves as a primary degradation signal characterizing the increase in the current spectral disorder.

[0010] Optionally, the thermal fatigue echo template convolution operation specifically includes: constructing a thermal fatigue echo template that characterizes the typical spectral features of thermal fatigue damage; performing a convolution operation between the current first-order derivative vector corresponding to the primary degradation signal and the thermal fatigue echo template vector to obtain a convolution result sequence; extracting echo segments in the convolution result sequence whose energy exceeds a preset energy threshold, and calculating the sum of squares of the convolution results within the echo segments as the echo energy value; the echo energy value reflects the matching energy between the current spectral feature and the fatigue damage template, and serves as the intermediate degradation signal.

[0011] Optionally, the negative step sharpness calculation specifically includes: real-time monitoring of the changes in the output intermediate degradation signal on a sub-second time scale; when a falling edge transition is detected in the intermediate degradation signal value, it is identified as a negative step event; calculating the average absolute value of the slope of the negative step event within a short time window; the average absolute value of the slope is used as the final degradation signal characterizing the suddenness and severity of the degradation process.

[0012] Optionally, S2 further includes weighted fusion of the primary degradation signal, intermediate degradation signal, and final degradation signal obtained from the calculation to calculate the final degradation index.

[0013] Optionally, the bivariate conjugate determination includes pre-setting a normal temperature range threshold and a degradation index threshold for the measured point; at each determination time, synchronously acquiring the current temperature time series data point and the corresponding real-time degradation index; comparing the temperature time series data point with the normal temperature range threshold, and simultaneously comparing the real-time degradation index with the degradation index threshold.

[0014] Optionally, the bivariate conjugate determination further includes determination and labeling based on the combination relationship of the bivariate comparison results: If the temperature time series data point is within the normal temperature range and the real-time degradation index is greater than the degradation index threshold, then the data status at that moment is marked as a Type I latent anomaly. If the temperature time series data point exceeds the normal temperature range, and the real-time degradation index is not greater than the degradation index threshold, then the data status at that moment is marked as a Type II explicit anomaly. If the temperature time series data point exceeds the normal temperature range, and the real-time degradation index is greater than the degradation index threshold, then the data status at that moment is marked as a Type III coupling anomaly. The Class I, Class II, and Class III anomaly marking results will serve as the logical basis for triggering subsequent differentiated alarms or protection control actions.

[0015] The beneficial effects of this invention are: This invention, through a bandpass spectral shape extraction mechanism, transforms the original reflection spectrum into a first-order derivative vector independent of temperature physical quantities, forming a pure shape sequence, effectively shielding the interference of factors such as light source power fluctuations and coupling losses. Based on this, through spectral entropy increase calculation, it realizes a quantitative assessment of the disorder of spectral shape, enabling the early identification of implicit degradation trends such as internal stress evolution and interface mismatch in materials before the temperature exceeds the limit. Compared with the traditional method that only relies on temperature threshold judgment, it has higher sensitivity.

[0016] This invention constructs a thermal fatigue echo template vector, performs convolution with the current first-order derivative spectral sequence, and extracts high-matching energy segments to achieve quantitative identification of characteristic perturbation patterns during degradation. Furthermore, it monitors the negative step change of degradation energy within a sub-second time window and extracts the average slope as a dynamic feature of sudden degradation, thereby forming a fusion model of three types of degradation signals: entropy increment, matching energy, and abrupt change sharpness. This improves the accuracy of identifying structural fatigue evolution and sudden deterioration processes.

[0017] This invention introduces a dual-variable cross-judgment logic of temperature time series and degradation index, combined with preset normal temperature range and degradation threshold, to achieve classification and identification of Type I latent anomalies, Type II explicit anomalies, and Type III coupled anomalies. This mechanism not only retains the monitoring capability for direct faults such as overheating, but also can mark potential risks in advance for which the degradation signal has increased significantly even though the temperature has not exceeded the limit. This provides a more refined judgment basis for subsequent differentiated alarm level control, redundancy switching strategy, and protection logic triggering. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Fig. 1 This is a schematic diagram of the identification method flow according to an embodiment of the present invention; Fig. 2 This is a schematic diagram of the shape-degradation mapping framework according to an embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0021] like Figs. 1-2 As shown, a method for identifying abnormal temperature data based on gallium arsenide fiber optic temperature measurement includes the following steps: S1, the temperature time series and reflection spectrum sequence of the measured point are synchronously acquired by the gallium arsenide temperature measurement fiber optic sensor, and the bandpass spectrum shape of the reflection spectrum sequence is extracted to obtain a pure shape sequence decoupled from the temperature physical quantity. The pure shape sequence is the first derivative vector of the reflection spectrum at the same time after normalization.

[0022] In gallium arsenide (GaAs) fiber optic temperature measurement (S11), the fiber itself does not spontaneously generate measurement data; it must rely on an external excitation source, namely a laser. The laser emits a series of laser beams of different wavelengths, called scanning light. This scanning light is fed into a GaAs fiber optic sensor. As the light propagates through the fiber, it encounters refractive index changes, impurities, or internal structures at the interface between the GaAs crystal and the fiber, resulting in partial reflection. This reflection varies with temperature, manifesting as changes in reflected light intensity and drift / deformation of the reflected wavelength. These reflected light signals are simultaneously received and recorded chronologically. The reflection signal at each moment is called a reflection spectrum, essentially a set containing multiple wavelength points and their corresponding reflection intensity values.

[0023] A broadband tunable laser continuously emits wavelengths ranging from [wavelength range missing] to a gallium arsenide temperature sensing fiber optic sensor. The scanning light signal is received synchronously, and its reflected signal is also received: The incident wavelength sequence is as follows: ; Indicates the first Each wavelength sampling point; Each time point The corresponding reflection spectrum is denoted as: ;in Indicates time Time Wavelength The intensity of reflected light.

[0024] In practice, the scanning wavelength range of a broadband tunable laser needs to be selected based on the spectral response characteristics of gallium arsenide (GaAs). GaAs exhibits high optical stability and usable reflection response in the near-infrared region, making it particularly suitable for high-temperature or long-distance temperature measurement scenarios. This avoids the impact of excessive absorption in the 1350–1450 nm range on measurement accuracy. The recommended wavelength scanning range is 1290 nm–1610 nm. This range covers the temperature-sensitive band, avoids major interference areas, and is compatible with mainstream GaAs reflection response characteristics.

[0025] S12, Channel Decomposition and Temperature / Spectral Data Acquisition: The received reflected signal is decomposed into two processing channels: Intensity demodulation channel: The purpose of this channel is to extract the specific temperature value from the reflected signal. First, the reflected light is sent to a photodetector, which converts it into an analog or digital signal representing the light intensity value at each wavelength. In other words, the photodetector converts the reflected light into an electrical signal. The reflection intensity at different wavelengths in the fiber optic reflection spectrum... Unlike its sensitivity to temperature changes, the reflection intensity within a certain wavelength band is weighted according to certain weights, and expressed as follows: ;in The temperature demodulation weighting factor follows the principle of maximizing temperature sensitivity, prioritizing wavelengths that show a significant response to temperature changes with higher weights. For example, if the reflection intensity is found to be most strongly correlated with temperature around 1480 nm, this band will be assigned a higher weight, while other bands will have lower weights. It is the first At the 1st time point, the 1st The weighted intensity is obtained by taking the reflection intensity of each wavelength. Then, it will be input into a temperature calibration model, and after being solved by a linear temperature calibration model, a continuous temperature time series will be obtained: ;in, For the first The temperature values ​​at each instant. The temperature calibration model is established based on experimental data, recording the weighted reflection intensity at each temperature point under known ambient temperature. Then, a set of temperature-intensity correlation equations is fitted. The solution process includes: collecting the reflection intensity at the new moment. Calculate the weighted sum ,Will Substitute into the model to calculate the corresponding .

[0026] Spectral Demodulation Channel: This channel's task is to acquire and retain the complete reflectance spectrum at each time point for subsequent spectral shape analysis. Using a spectrometer, the reflected light at each moment is broken down by wavelength and recorded as a light intensity sequence containing multiple wavelength points, called the reflectance spectrum. This reflectance spectrum data will be used for subsequent spectral shape derivative extraction and degradation index identification. In other words, it is acquired by the spectrometer. The reflection spectrum sequence that forms the time index.

[0027] S13, Reflectance Bandpass Extraction and Pure Shape Construction: At each time point, a reflection spectrum is received from the gallium arsenide fiber optic temperature sensor. This reflection spectrum is a light intensity distribution arranged by wavelength, containing hundreds or even thousands of wavelength sampling points. Essentially, it represents the reflection intensity of light at different wavelengths at a given moment. However, this raw spectral signal simultaneously contains temperature-related spectral shape drift or deformation, global intensity changes related to light source power fluctuations, coupling efficiency, and environmental disturbances. To extract the spectral shape information affected only by microscopic changes in the material's internal structure, for each reflection spectrum... Perform the following operations to extract the spectral information of the decoupled temperature physical quantity: S131, Selecting a Bandpass Window: Since the reflection spectrum typically covers a wide wavelength range, only certain local bands are more sensitive to temperature-related internal structural changes, such as lattice vibrations and changes in doping states, while other bands may be insensitive to temperature changes or have high noise levels. Therefore, it is necessary to select a reference bandpass window from the entire spectrum, retaining only a portion of the wavelengths within this window as the object of analysis. The selection of this window serves the following purposes: focusing on the most stable and representative part of the spectral shape, and shielding spectral bands that are meaningless for temperature measurement; that is, setting a reference spectral shape bandpass window based on the temperature-sensitive peak positions of gallium arsenide material. The spectral subsequence corresponding to this window is: ; Indicates the first The reflection spectrum sequence within the bandpass window at each time point; When selecting a bandpass window, the spectral response characteristics of gallium arsenide (GaAs) materials need to be considered comprehensively. GaAs has stable reflection characteristics in the near-infrared region. In the range of 1420–1510 nm, its reflection spectrum is highly sensitive to structural changes, but its sensitivity to temperature changes is relatively weak. Therefore, it is suitable as a stable reference window. The water absorption peak near 1350–1400 nm should also be avoided. The range of values ​​should be 1430 nm–1490 nm.

[0028] S132. Due to factors such as light source output power and fiber optic connection loss, the overall intensity of the reflected spectrum may fluctuate with each acquisition. Therefore, the absolute value of the intensity cannot be directly used for analysis. To address this, the reflected intensity within the selected bandpass window needs to be normalized to a minimum-maximum value. This means compressing all intensity values ​​within the window to the range [0,1] to eliminate the influence of absolute intensity variations and emphasize the local trend of the spectrum rather than the overall brightness. The minimum-maximum normalization is performed on the reflected intensity within the bandpass window as follows: ; S133, First derivative spectral shape extraction: The normalized light intensity sequence is then subjected to discrete first-order derivative calculation, which calculates the rate of change of light intensity between two adjacent wavelength points, forming a new vector sequence. This vector reflects whether the shape of the reflection spectrum steepens upwards, decreases gradually, or whether there are abrupt changes within a certain small waveband. The sequence constructed in this way is called a pure shape sequence, which is detached from the absolute value of the temperature itself and only reflects whether there are changes or degradations in the material's microscopic optical response to laser.

[0029] Calculate the discrete first derivative of the normalized light intensity sequence with respect to wavelength to obtain the spectral shape change rate vector: ; in This represents the number of sampling points within the pass-through window. This is the pure shape sequence corresponding to that moment, representing the local slope of the spectral shape.

[0030] S2, input the pure shape sequence into a preset shape-degradation mapping framework and output a degradation index. The shape-degradation mapping framework includes sequentially connected spectral entropy increase operation, thermal fatigue echo template convolution operation and negative step sharpness operation, and the value of the degradation index increases monotonically with the increase of spectral disorder, echo energy and negative step slope. 1. Increased spectral entropy reflects the overall trend of disorder in spectral shape. Each reflection spectrum is essentially a spectral shape distribution. In a healthy material state, this shape is relatively regular, smooth, and concentrated. When the internal structure of the material begins to undergo microscopic degradation, such as changes in lattice stress and defect density, the reflection spectrum gradually becomes more disordered, exhibiting enhanced local perturbations, misaligned peaks and valleys, and an overall tendency towards disorder. This degree of disorder can be quantified by calculating the information entropy of the first derivative vector. If the information entropy continues to increase, it indicates that the spectral shape is becoming more irregular, a statistical sign of early degradation. This is a generalized perception of the degradation trend.

[0031] 2. Template convolution identification and similarity to known fatigue patterns. Some specific degradation processes, such as thermal fatigue and microcrack initiation, exhibit fixed perturbation patterns in the spectrum, such as locally steeply rising and falling wave packets. If a structure similar to these templates appears in the current spectral derivative sequence, it indicates that a change is occurring that highly matches the structure of past damage patterns. The convolution operation detects the degree of this pattern matching, and the convolution energy reflects the strength of this matching. This is a direct extraction of degradation morphological features and has stronger structural indicative value, similar to the identification of abnormal waveforms in an electrocardiogram.

[0032] 3. Negative step sharpness captures sudden changes in the degradation process. Even if the degradation trend and morphology are not yet obvious, a sudden drop in a certain signal within a short period of time is very likely to represent a sudden break point or instability event. For example, if a region has just completed fatigue accumulation and then undergoes a structural abrupt change, the corresponding spectral signal will show a jump or sharp change at the sub-second level. Here, a negative step is like finding a collapse in a continuous signal, and the absolute value of its slope quantifies the severity of this sudden change. This is for monitoring the rate and severity of degradation.

[0033] This step constructs three sub-features—spectral disorder, thermal fatigue template matching energy, and abrupt degradation sharpness—using a shape-degradation mapping framework. These are then combined under a unified architecture to form multi-layered degradation signals, which are ultimately weighted and fused to generate degradation index values. These include the following: S21, Spectral Entropy Increase Operation: The goal is to assess whether the current spectral shape is becoming more disordered and chaotic by calculating the information entropy of the first derivative vector of the spectral shape, thus serving as an early signal of degradation. When a material has not yet significantly degraded, its spectral shape is usually relatively smooth, with only a few obvious slope changes. However, when the internal structure of the material begins to undergo microcracks, stress concentration, or changes in doped states, the spectral shape becomes more complex, fluctuating, asymmetrical, or chaotic. These changes manifest in the first derivative sequence as abrupt slope changes, reverse perturbations, and increased fluctuation frequencies. Specifically, this includes the input pure shape sequence, i.e., the... The first derivative vector at time t. Perform information entropy calculation: S211: First, the absolute value of each component in the pure shape sequence (first derivative vector) acquired at the current time point is normalized so that their sum is 1. This is equivalent to treating each small slope change as a probability event, representing the proportion of intensity where the spectrum undergoes a sudden change at this wavelength. In other words, normalizing the vector yields the probability distribution. ,in: ;in, Indicates the first In the pure shape sequence at time point n, the th One element, Describes the probability distribution of the th One component; S212, Calculate information entropy : ;in, To prevent extremely small positive numbers from having a logarithmic value of zero, It represents the total number of wavelength sampling points in each reflection spectrum; information entropy measures the degree of disorder in a distribution. If the changes are concentrated in a few bands, the entropy value is low (orderly shape); if the changes occur in many bands, the entropy value is high (chaotic shape).

[0034] S213, Set the baseline information entropy under historical normal conditions. The entropy increment, i.e., the primary degeneracy signal, is defined as: ; This represents the baseline value of spectral information entropy obtained based on historical normal data statistics; S22, Thermal fatigue echo template convolution operation: The goal is to identify whether the current spectrum shows a typical spectral perturbation pattern similar to that in the thermal fatigue damage process, and to quantify the degree of similarity with a numerical value.

[0035] In actual material degradation processes, the reflectance spectrum often exhibits regular abrupt changes, such as sharp peaks or valleys that suddenly rise and then rapidly decline within a certain wavelength range. These features can be extracted and constructed into a standard template vector, also known as a thermal fatigue template, reflecting the spectral abrupt change shape of a typical damage. If the current spectral derivative shape is very similar to this template, it means that degradation behavior corresponding to that template is occurring.

[0036] Specifically, this includes constructing a thermal fatigue echo template vector. This reflects the shape characteristics of spectral mutations under common degradation states. The first element in the thermal fatigue echo template vector represents the... Each template component is the length of the thermal fatigue echo template vector.

[0037] S221, for the first The first derivative vector at time t. With template Perform convolution: Where * denotes one-dimensional discrete convolution, The sequence of convolution results represents the nth... The result sequence obtained by convolving the first derivative vector at each time step with the thermal fatigue template. Represents the sequence of convolution results The Middle Each convolutional output value The length of the convolution result sequence is given. The essence of convolution is to continuously align two signal segments within a sliding window and observe their similarity. If the shape of a certain derivative closely matches the shape of the template, the corresponding convolution value will be high, ultimately resulting in a convolution result sequence that reflects which bands the shape of the current spectrum most closely resembles the fatigue template.

[0038] The thermal fatigue echo template is a fixed one-dimensional vector, and the construction process is as follows: 1. Select multiple test samples that show fatigue cracks or degradation after long-term thermal cycling experiments; statistically analyze the reflection spectrum derivatives recorded in the early stage of degradation of these samples, and select those representative spectrum shapes with local severe fluctuations, asymmetric slopes, etc. as candidates.

[0039] 2. Unify the length of the selected derivative segments, and perform min-max normalization or Z-score standardization on each vector segment to ensure that the feature scale is consistent among different samples.

[0040] 3. Perform point alignment averaging on all standardized candidate spectra to obtain a stable central trend. Use smoothing filtering to eliminate individual noise and enhance representativeness, ultimately forming a standard template vector.

[0041] S222, Set the convolution energy threshold. Used to screen significant echo segments and extract those that meet the criteria. section And calculate the convolution energy of this segment, i.e., the intermediate degradation signal: Convolution energy threshold Based on the statistical distribution of historical normal samples, the absolute values ​​of all convolution results were statistically analyzed from a large amount of spectral derivative data under healthy conditions, and the 95th percentile value was used as the upper limit of normal fluctuation.

[0042] S23, Negative Step Sharpness Calculation: In the degradation process of some materials or structures, especially in the early stages of thermal fatigue or microcrack propagation, the changes in the reflection spectrum are often sudden rather than stable and continuous. That is to say, they suddenly deteriorate at a certain moment, manifested as a rapid jump or sudden drop in the characteristic signal (the intermediate degradation signal obtained by convolution). This kind of abrupt change is called a negative step event.

[0043] Specifically, this includes monitoring intermediate degradation signal sequences at a time resolution of approximately 100ms. Identify whether there is a significant falling edge transition, i.e., a negative step event: S231, if it exists Determine the moment There is a negative step. The value of the intermediate degradation signal from the previous moment. The negative step threshold represents the numerical limit for determining whether a sudden jump is significant. It is calculated by statistically analyzing the distribution of consecutive point differences from a large number of intermediate degradation signal sequences in normal or non-mutated states. ; This represents the average signal difference under normal conditions. Standard deviation An empirical coefficient is set between 2.0 and 2.5; if a drop exceeds this threshold, it is considered an abnormal drop.

[0044] S232, after confirming the occurrence of a negative step, back one time window from the current moment and observe the rate of change of the intermediate degradation signal within this short period. This includes calculating the slope of the signal change between each pair of adjacent sampling points within this time period, averaging the absolute values ​​of these slopes to obtain an index representing the degree of drastic change. That is, within the time window... Within this process, the average of the absolute values ​​of the slopes is calculated as the final degradation signal: ;in, Indicates the first The time value at each sampling moment This indicates the number of historical points contained within a sub-second time window. Indicates the current time point Backtracking The time point after one sampling period, that is... This is the starting moment of the current sub-second time window. Indicates the first Intermediate degradation signal values ​​at each time point Indicates the first Sampling time at each time point; S24, Degradation Index Synthesis: The degradation signals at three levels are weighted and fused to obtain the comprehensive degradation index at the current moment. : ;in , which are the fusion weights of the three types of signals, and can be set empirically or optimized through training based on actual data.

[0045] S3. Perform bivariate conjugate determination on the temperature time series and degradation index. When the temperature time series is within the normal temperature range and the degradation index is greater than the degradation threshold, it is marked as a Type I latent anomaly. When the temperature time series exceeds the normal temperature range and the degradation index is not greater than the degradation threshold, it is marked as a Type II explicit anomaly. When the temperature time series exceeds the normal temperature range and the degradation index is greater than the degradation threshold, it is marked as a Type III coupled anomaly. The Type I, Type II, and Type III marking results are used to trigger subsequent differentiated protection actions.

[0046] At each judgment moment It simultaneously acquires the temperature value and corresponding degradation index value of the target measurement point, and performs bivariate conjugate judgment based on the preset normal temperature range and degradation index threshold to achieve data status classification and label assignment. This includes the following: S31, Threshold setting: Two key decision thresholds are predefined: Normal temperature range: ; Degradation index threshold: ; in, Indicates the first Temperature time-series values ​​collected at each time point Indicates the first The degradation index value corresponding to each time point This represents the normal temperature range. Based on historical operating data statistics, temperature time series are extracted from long-term historical data, and abnormal and operating condition switching data are removed. The 95% confidence interval or the interval between the upper and lower means ±2σ is calculated for the steady-state operating period. ;in The average temperature. Standard deviation; Let D(t) represent the abnormal threshold value of the degradation index. This is determined by observing the lower limit of the numerical distribution of D(t) from historical data of multiple clearly abnormal states. This ensures that the system generates an out-of-limit response in the early stages of degradation.

[0047] S32, Decision Logic and Status Flags: Based on the bivariate relationship between temperature and degradation indicators, the system executes the following three types of decision logic: S321, Type I latent abnormality (potential early regression): Marked as: Type I latent anomaly; S322, Type II dominant abnormality (temperature abnormality only): and Marked as: Industry-specific dominant anomaly; S323, Type III Coupling Anomaly (Temperature + Degradation Dual Anomaly): and Marked as: Type III coupling anomaly.

[0048] S33, the result of each type of anomaly labeling will be input into the subsequent control unit of the system to trigger different levels of alarm or protection response: Category I: Issues a warning signal and recommends diagnosis; Category II: Activate environmental anomaly handling strategies; Class III: Initiate high-level linkage protection, including temperature control power failure, redundancy switching, etc.

[0049] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0050] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry, characterized in that, Includes the following steps: S1, the temperature time series and reflection spectrum sequence of the measured point are synchronously acquired by the gallium arsenide temperature measurement fiber optic sensor, and the bandpass spectrum shape of the reflection spectrum sequence is extracted to obtain a pure shape sequence decoupled from the temperature physical quantity. The pure shape sequence is the first derivative vector of the reflection spectrum at the same time after normalization. S2, input the pure shape sequence into a preset shape-degradation mapping framework and output a degradation index. The shape-degradation mapping framework includes sequentially connected spectral entropy increase operation, thermal fatigue echo template convolution operation and negative step sharpness operation, and the value of the degradation index increases monotonically with the increase of spectral disorder, echo energy and negative step slope. S3. Perform bivariate conjugate judgment on temperature time series and degradation index. When the temperature time series is in the normal temperature range and the degradation index is greater than the degradation threshold, it is marked as a Type I latent anomaly. When the temperature time series exceeds the normal temperature range and the degradation index is not greater than the degradation threshold, it is marked as a type II overt abnormality; When the temperature time series exceeds the normal temperature range and the degradation index is greater than the degradation threshold, it is marked as a Type III coupling anomaly. The Class I, Class II, and Class III marking results are used to trigger subsequent differentiated protection actions.

2. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 1, characterized in that, The S1 further includes transmitting scanning light to the gallium arsenide temperature sensing fiber optic sensor via a broadband tunable laser and simultaneously receiving its reflected light signal.

3. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 2, characterized in that, The reflected light signal is decomposed into an intensity demodulation channel and a spectral demodulation channel: The intensity demodulation channel is converted into an electrical signal by a photodetector, and then calculated by a temperature calibration model to obtain continuous temperature time series data; The spectral demodulation channel acquires a time-indexed reflectance spectrum sequence through a spectrometer, with each reflectance spectrum including light intensity values ​​at multiple wavelength sampling points.

4. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 3, characterized in that, Bandpass spectral shape extraction is performed on each reflection spectrum in the reflection spectrum sequence: Select a reference spectral bandpass window that is adjacent to the temperature-sensitive peak of gallium arsenide material and is insensitive to temperature changes; The reflectance spectrum data within the reference spectral bandpass window is extracted and subjected to minimum-maximum normalization to eliminate the influence of fluctuations in the absolute value of light intensity. Calculate the first derivative of the minimum-maximum normalized spectrum to form a vector describing the local rate of change of the spectrum. This vector is the pure shape sequence decoupled from the temperature physical quantity.

5. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 1, characterized in that, The spectral entropy increase operation specifically includes: calculating the information entropy of the first derivative vector as the initial disorder value; comparing the initial disorder value at the current moment with a benchmark entropy value obtained from historical normal data statistics, and calculating its relative increment, wherein the relative increment serves as a primary degradation signal characterizing the increase in the current spectral disorder.

6. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 5, characterized in that, The thermal fatigue echo template convolution operation specifically includes: constructing a thermal fatigue echo template that characterizes the typical spectral features of thermal fatigue damage; performing a convolution operation between the current first-order derivative vector corresponding to the primary degradation signal and the thermal fatigue echo template vector to obtain a convolution result sequence; extracting echo segments in the convolution result sequence whose energy exceeds a preset energy threshold, and calculating the sum of squares of the convolution results within the echo segments as the echo energy value; the echo energy value reflects the matching energy between the current spectral feature and the fatigue damage template, and serves as the intermediate degradation signal.

7. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 6, characterized in that, The negative step sharpness calculation specifically includes: real-time monitoring of the changes in the output intermediate degradation signal on a sub-second time scale; when a falling edge transition is detected in the intermediate degradation signal value, it is identified as a negative step event; the average absolute value of the slope of the negative step event within a short time window is calculated; the average absolute value of the slope is used as the final degradation signal characterizing the suddenness and severity of the degradation process.

8. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 1, characterized in that, S2 further includes weighted fusion of the primary degradation signal, intermediate degradation signal, and final degradation signal obtained from the calculation to calculate the final degradation index.

9. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 1, characterized in that, The bivariate conjugate determination includes pre-setting a normal temperature range threshold and a degradation index threshold for the measured point; at each determination time, the temperature time series data point and the corresponding real-time degradation index are acquired synchronously; the temperature time series data point is compared with the normal temperature range threshold, and the real-time degradation index is compared with the degradation index threshold.

10. The method for identifying abnormal temperature data based on gallium arsenide fiber optic thermometry according to claim 9, characterized in that, The bivariate conjugate determination also includes determination and labeling based on the combination relationship of the bivariate comparison results: If the temperature time series data point is within the normal temperature range and the real-time degradation index is greater than the degradation index threshold, then the data status at that moment is marked as a Type I latent anomaly. If the temperature time series data point exceeds the normal temperature range, and the real-time degradation index is not greater than the degradation index threshold, then the data status at that moment is marked as a Type II explicit anomaly. If the temperature time series data point exceeds the normal temperature range, and the real-time degradation index is greater than the degradation index threshold, then the data status at that moment is marked as a Type III coupling anomaly. The Class I, Class II, and Class III anomaly marking results will serve as the logical basis for triggering subsequent differentiated alarms or protection control actions.

Images