Hybrid optical fiber sensing method for temperature monitoring of water turbine generator rotor lead

Abstract

The present application relates to the technical field of motor group state monitoring, and discloses a kind of water turbine generator rotor lead temperature monitoring methods fused optical fiber sensing, comprising the following steps, in stator side deployment integrated fluorescence temperature measurement, frequency-modulated continuous wave ranging and spectrum acquisition unit multimodal optical detection assembly, establish full circumferential space-time synchronization benchmark;Trigger detection assembly to execute global inspection, synchronous acquisition fluorescence decay, frequency-modulated continuous wave beat frequency and background spectrum signal;Demodulation signal and reconstruct one-dimensional temperature field, three-dimensional geometric vibration field and chemical fingerprint field;Lock target area based on temperature anomaly, utilize measured distance correction two-dimensional galvanometer deflection parameter, execute active tracking and local high-resolution scanning;Extract thermal, mechanical and chemical multidimensional feature vector, input multi-physical quantity correlation model output fault type.The present application realizes the accurate identification and positioning of mechanical looseness, insulation breakdown and abnormal heat dissipation in high-speed rotating environment through multi-source feature fusion.

Description

Technical Field

[0001] This invention relates to the field of generator set condition monitoring technology, specifically a method for monitoring the rotor lead temperature of a hydro-generator that integrates fiber optic sensing. Background Technology

[0002] The rotor leads and current collector rings of a hydro-generator unit are critical channels for the introduction of excitation current into the rotor windings. During operation, they are subjected to high voltage, high current, and enormous centrifugal forces from high-speed rotation. Due to the harsh operating environment, this area is prone to faults such as loose connections, increased contact resistance, insulation degradation, and even partial discharge. Once a fault occurs, it is often accompanied by a dramatic temperature rise; if not detected and addressed promptly, it can lead to burn-out accidents, seriously threatening the safe and stable operation of the unit. Therefore, real-time and accurate condition monitoring of the rotor leads is of paramount importance.

[0003] Temperature monitoring of rotating components primarily employs two types of technologies: contact measurement and non-contact measurement. Contact measurement typically uses wireless temperature sensors or slip rings to extract signals. However, these sensors must be mounted on the surface of the high-speed rotating rotor, which not only makes installation and maintenance difficult but may also disrupt the rotor's dynamic balance. Furthermore, the strong electromagnetic field environment inside the generator can easily interfere with wireless signal transmission, leading to data loss or distortion. Non-contact measurement often uses infrared thermal imaging technology. While avoiding physical contact, it has significant limitations in practical applications. The metallic luster of the rotor lead surface results in low and unstable emissivity, severely affecting temperature measurement accuracy. Moreover, under high-speed rotation conditions, traditional infrared cameras are limited by frame rate and integration time, often resulting in blurred or trailing images, making it difficult to capture overheating characteristics in minute areas.

[0004] Existing monitoring methods are mostly limited to detecting a single physical quantity, temperature. However, the causes of rotor lead wire faults are complex. Poor contact due to loose mechanical connections, partial discharge caused by insulation damage, and heat dissipation obstruction due to oil contamination, while all ultimately manifesting as temperature rise, have drastically different fault mechanisms. Relying solely on temperature data cannot accurately distinguish specific fault types, making it difficult for maintenance personnel to determine the root cause. For example, mechanical loosening is often accompanied by minute radial vibration changes, partial discharge ionizes the surrounding air to produce ozone, and oil accumulation alters the spectral characteristics of the surface. Current technology lacks a comprehensive monitoring solution capable of simultaneously sensing multi-dimensional thermal, mechanical, and chemical characteristics and accurately locating and tracking high-speed rotating targets, resulting in a high false alarm rate in fault diagnosis and failing to meet the needs of intelligent operation and maintenance of generator sets. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for monitoring the temperature of the rotor leads of a hydro-generator that integrates fiber optic sensing. This method solves the problem that existing generator rotor lead monitoring technologies are limited in scope and cannot simultaneously acquire temperature, vibration, and environmental chemical characteristics under high-speed rotation and strong electromagnetic environments, thus making it impossible to accurately distinguish between different fault types such as mechanical loosening, insulation degradation, and heat dissipation obstruction.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for monitoring the temperature of the rotor leads of a hydro-generator integrating fiber optic sensing, comprising the following steps:

[0007] S1. Construct a multimodal optical sensing environment and establish a spatiotemporal synchronization reference; Deploy a multimodal optical detection assembly integrating a fluorescence temperature measurement unit, a frequency-modulated continuous wave ranging unit, a spectrum acquisition unit and a two-dimensional scanning galvanometer on the generator stator side; Prepare a fluorescence-matrix composite functional coating on the rotor lead surface; Establish a full-circumferential spatiotemporal synchronization reference using a position encoder.

[0008] S2. Based on the spatiotemporal synchronization reference, the multimodal optical detection assembly is triggered to perform a global inspection, and fluorescence attenuation signal, frequency-modulated continuous wave beat frequency signal and environmental background spectrum signal are simultaneously collected at multiple discrete sampling angle positions.

[0009] S3. Demodulate the original physical quantity signal and reconstruct the multidimensional state field; demodulate the fluorescence attenuation signal in the time domain to reconstruct the one-dimensional circumferential temperature distribution field; analyze the frequency domain of the frequency-modulated continuous wave beat signal to reconstruct the three-dimensional geometric contour and vibration state field; integrate the characteristic peaks of the environmental background spectral signal to reconstruct the environmental chemical fingerprint distribution field.

[0010] S4. Compare the one-dimensional circumferential temperature distribution field with a preset threshold. When an anomaly is detected, lock the center angle coordinates according to the spatiotemporal synchronization reference and trigger active tracking and local scanning modes.

[0011] S5. Calculate the geometric correction deflection parameters of the two-dimensional scanning galvanometer based on the instantaneous distance value corresponding to the center angle coordinate in the three-dimensional geometric contour and vibration state field, and drive the multi-modal optical detection assembly to perform grating scanning on the abnormal area to generate a two-dimensional high-resolution temperature matrix and a two-dimensional spectral fingerprint matrix.

[0012] S6. Extract multidimensional fault feature vectors containing thermal, mechanical and chemical properties from the one-dimensional circumferential temperature distribution field, the three-dimensional geometric contour and vibration state field, the two-dimensional high-resolution temperature matrix and the two-dimensional spectral fingerprint matrix; input the multidimensional fault feature vectors into a preset multi-physical quantity correlation diagnostic model, and output the fault type and location result of the rotor lead.

[0013] Preferably, in step S1, the fluorescent-matrix composite functional coating is prepared by mixing rare earth ion-doped inorganic phosphor with a high reflectivity ceramic matrix material and then using an atmospheric plasma spraying process. The high reflectivity ceramic matrix material is alumina or zirconium oxide.

[0014] The multimodal optical detection assembly adopts a common optical path design, wherein the two-dimensional scanning galvanometer is located at the beam exit port of the multimodal optical detection assembly, and is used to uniformly deflect and control the detection beams of the fluorescence thermometry unit, the frequency-modulated continuous wave ranging unit and the spectral acquisition unit.

[0015] Preferably, in step S2, triggering the multimodal optical detection assembly based on the spatiotemporal synchronization reference includes:

[0016] The mechanical angle of one revolution of the rotor is divided into multiple discrete sampling sectors, and a unique angle index is assigned to each sampling sector.

[0017] A spatial trigger event is generated when the real-time angle value fed back by the position encoder matches any sampled angle position;

[0018] In response to the spatial triggering event, the fluorescence thermometry unit is simultaneously driven to emit a pulsed laser, the frequency-modulated continuous wave ranging unit is driven to start a linear frequency-modulated sweep cycle, and the electronic shutter of the spectral acquisition unit is opened, so as to acquire the fluorescence attenuation signal, the frequency-modulated continuous wave beat frequency signal, and the environmental background spectral signal in parallel within the same time window.

[0019] Preferably, step S3, which involves demodulating the original physical quantity signal and reconstructing the multidimensional state field, specifically includes:

[0020] The fluorescence decay signal is fitted with an exponential decay model using the nonlinear least squares method to obtain the fluorescence lifetime constant, and the absolute temperature value is calculated based on the calibrated temperature-lifetime characteristic curve to construct the one-dimensional circumferential temperature distribution field.

[0021] Perform a fast Fourier transform on the frequency-modulated continuous wave beat frequency signal, extract the main frequency component to calculate the probe optical path, and combine the rotor rotation radius to convert the distance data in the polar coordinate system into the rectangular coordinate system data to construct the three-dimensional geometric profile and vibration state field;

[0022] For the target chemical component to be monitored, a characteristic response band range is preset. The environmental background spectral signal is integrated within the characteristic response band range and the baseline intensity is subtracted to calculate the spectral characteristic intensity index and construct the environmental chemical fingerprint distribution field.

[0023] Preferably, step S4, which involves locking the center angle coordinates of the abnormal region and triggering active tracking and local scanning modes, includes:

[0024] Calculate the difference between the temperature data in the one-dimensional circumferential temperature distribution field and the preset temperature alarm threshold. When the difference is greater than zero, generate an abnormal interruption signal.

[0025] Peak search is performed on the continuous over-temperature range to determine the center angle coordinates of the abnormal region;

[0026] Based on the current instantaneous angular velocity and current rotor angle fed back by the position encoder, the predicted arrival time of the center angle coordinate rotating to the optical axis alignment position of the multimodal optical detection assembly is calculated;

[0027] Based on the predicted arrival time, the scheduling system enters active tracking mode and preloads the control data of the two-dimensional scanning galvanometer.

[0028] Preferably, the calculation of the geometric correction deflection parameters of the two-dimensional scanning galvanometer in step S5 specifically involves:

[0029] The measured vertical distance corresponding to the center angle coordinates, stored in the three-dimensional geometric contour and vibration state field, is called as the reference distance parameter;

[0030] For any pixel within a local scanning area, obtain its horizontal and vertical physical offsets relative to the scanning center.

[0031] Calculate the ratio of the lateral physical offset to the measured vertical distance, and take the arctangent of this ratio as the X-axis galvanometer deflection angle;

[0032] Calculate the ratio of the longitudinal physical offset to the measured vertical distance, and take the arctangent of this ratio as the Y-axis galvanometer deflection angle;

[0033] The voltage sequence generated by the X-axis galvanometer deflection angle and the Y-axis galvanometer deflection angle drives the two-dimensional scanning galvanometer to eliminate imaging distortion caused by changes in rotor distance.

[0034] Preferably, the generation of the two-dimensional high-resolution temperature matrix and the two-dimensional spectral fingerprint matrix in step S5 includes:

[0035] Within the active tracking time window, the two-dimensional scanning galvanometer is controlled to deflect according to a preset grating path;

[0036] During the scanning process, the fluorescence thermometry unit is used to obtain the local temperature value of each pixel on the scanning path, forming the two-dimensional high-resolution temperature matrix.

[0037] Simultaneously, the environmental spectral feature values ​​of each pixel on the scanning path are acquired using the spectral acquisition unit to form the two-dimensional spectral fingerprint matrix;

[0038] The two-dimensional high-resolution temperature matrix and the two-dimensional spectral fingerprint matrix are strictly corresponding in terms of spatial pixel index.

[0039] Preferably, the process of constructing the multidimensional fault feature vector in step S6 includes:

[0040] Extract the local maximum temperature value from the two-dimensional high-resolution temperature matrix and calculate the spatial gradient value of the temperature distribution;

[0041] Extract the radial vibration amplitude corresponding to the center angle coordinate from the three-dimensional geometric contour and vibration state field;

[0042] Extract the spectral intensity integral values ​​corresponding to the ozone characteristic band and the spectral intensity integral values ​​corresponding to the oil mist characteristic band from the two-dimensional spectral fingerprint matrix.

[0043] The local maximum temperature value, the spatial gradient value, the radial vibration amplitude, the spectral intensity integral value of the ozone characteristic band, and the spectral intensity integral value of the oil mist characteristic band are normalized and combined to form the multidimensional fault feature vector.

[0044] Preferably, in step S6, the multidimensional fault feature vector is input into a preset multi-physical quantity correlation diagnostic model, and the fault type of the rotor lead is output, including the following logical criteria:

[0045] When the local maximum temperature value is higher than the temperature threshold, and the radial vibration amplitude is higher than the vibration threshold, while the spectral intensity integral value of the ozone characteristic band and the spectral intensity integral value of the oil mist characteristic band are both lower than their respective spectral intensity thresholds, the fault type is determined to be loose mechanical connection or fastener fatigue.

[0046] When the local maximum temperature value is higher than the temperature threshold, the spatial gradient value is higher than the gradient threshold, and the spectral intensity integral value of the ozone characteristic band is higher than the spectral intensity threshold, the fault type is determined to be partial discharge or insulation layer breakdown.

[0047] When the local maximum temperature value is higher than the temperature threshold, but the spatial gradient value is lower than the gradient threshold, and the spectral intensity integral value of the oil mist characteristic band is higher than the spectral intensity threshold, the fault type is determined to be heat dissipation obstruction caused by oil stains.

[0048] Preferably, the method for calculating the radial vibration amplitude is as follows:

[0049] Maintain a sliding time window and statistically average the distance data within a preset number of rotation cycles in the past to obtain the static reference distance at the center angle coordinates;

[0050] Calculate the difference between the instantaneous distance of the current cycle and the static reference distance, and determine the absolute value of the difference as the radial vibration amplitude;

[0051] The instantaneous distance is calculated by the frequency-modulated continuous wave ranging unit according to the linear frequency modulation ranging principle. Specifically, it is the value obtained by dividing the product of the vacuum speed of light, the laser frequency modulation period, and the main frequency component of the beat frequency signal by twice the laser modulation bandwidth.

[0052] This invention provides a method for monitoring the temperature of the rotor leads of a hydro-generator that integrates fiber optic sensing. It has the following beneficial effects:

[0053] 1. This invention overcomes the limitations of traditional single-physical-quantity monitoring in fault tracing by constructing a multimodal sensing system that includes fluorescence thermometry, frequency-modulated continuous wave ranging, and spectral acquisition. The system can simultaneously acquire information on the thermal distribution, mechanical vibration state, and chemical composition of the surrounding environment of the rotor leads from the same monitoring point. Utilizing a multi-physical-quantity correlation diagnostic model, it effectively distinguishes different types of overheating faults caused by loose mechanical connections, partial discharge of the insulation layer, or accumulation of surface oil, significantly improving the specificity and accuracy of generator rotor fault diagnosis.

[0054] 2. This invention employs active tracking and local scanning technology based on a spatiotemporal synchronization reference, solving the problems of difficult fixed-point monitoring and low data spatial resolution under high-speed rotation conditions. Through the coordinated control of the position encoder and the two-dimensional scanning galvanometer, the system can automatically lock the target area after detecting a coarse anomaly, and use the real-time measured depth and distance data to perform geometric correction on the galvanometer deflection angle, eliminating imaging distortion caused by rotor radial displacement, thereby achieving microsecond-level precise tracking and high-resolution thermal-spectral imaging of high-speed moving targets.

[0055] 3. This invention constructs an intelligent diagnostic model based on the correlation of multiple physical quantities. It can accurately identify different types of faults, such as loose mechanical connections, insulation breakdown, or impeded heat dissipation due to oil buildup, based on a combination of temperature gradient, radial vibration amplitude, and characteristic spectral intensity. By normalizing multi-source heterogeneous data and performing logical correlation analysis, this method breaks through the information silo effect of traditional threshold alarm modes, directly outputting the specific fault type and location, significantly reducing the difficulty of troubleshooting and maintenance costs for operation and maintenance personnel. Attached Figure Description

[0056] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0057] The technical solutions in 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.

[0058] Please see the appendix Figure 1 This invention provides a method for monitoring the temperature of the rotor leads of a hydro-generator that integrates fiber optic sensing, comprising the following steps;

[0059] S1. Construct a multimodal optical sensing environment and establish a spatiotemporal synchronization reference; deploy a multimodal optical detection assembly on the generator stator side and prepare a fluorescent-matrix composite functional coating on the rotor lead surface; use a position encoder connected to the rotor shaft end to obtain the rotor rotation angle signal and establish a full-circumferential spatiotemporal synchronization reference; the multimodal optical detection assembly integrates a fluorescent temperature measurement unit, a frequency-modulated continuous wave ranging unit, a spectral acquisition unit, and a two-dimensional scanning galvanometer.

[0060] Specifically, step S1 in this embodiment includes the following sub-steps:

[0061] S110 was used to prepare a functional coating with dual optical properties. This fluorescent-matrix composite functional coating was applied to the surface of the rotor leads of the hydro-generator to be monitored, particularly at the busbar joints, lead connectors, and inter-electrode connections. As a signal carrier for non-contact measurement, this functional coating must simultaneously possess temperature sensitivity and broadband reflectivity.

[0062] The fluorescent-matrix composite functional coating is formed by mixing a fluorescent active material and a high-reflectivity ceramic matrix material in a preset mass ratio. The fluorescent active material is selected from rare-earth ion-doped inorganic phosphors, and the ceramic matrix material is selected from alumina or zirconium oxide powder, utilizing their high refractive index properties to form a diffuse reflective surface.

[0063] A mixed powder was deposited onto the rotor lead surface using an atmospheric plasma spraying process, with the coating thickness controlled between 100 and 300 micrometers. The composite structure ensures both the excitation efficiency of the fluorescence signal and provides a high-reflectivity background, allowing non-fluorescent light signals from the environment to be reflected back to the detection system via the coating surface, providing optical echoes with a sufficient signal-to-noise ratio for subsequent environmental spectral fingerprint analysis.

[0064] S120, deploy and calibrate the four-dimensional optical detection assembly. Install the four-dimensional optical detection assembly on the reserved observation hole or bracket inside the generator stator frame. The four-dimensional optical detection assembly adopts a common optical path or off-axis optical path design, integrating detection units of different bands and functions into the same physical housing to ensure that the measurement fields of each physical quantity coincide in space.

[0065] The four-dimensional optical detection assembly includes: a fluorescence excitation and acquisition unit, which includes a pulsed laser and a photomultiplier tube (PMT), and separates the excitation beam from the returned fluorescence signal through a dichroic mirror;

[0066] The frequency-modulated continuous wave ranging unit includes a tunable laser diode and a photodiode, used to emit and receive frequency-modulated continuous wave lasers;

[0067] The spectral acquisition unit, which includes a miniature spectrometer or a multi-band photodiode array, uses a beam splitter to extract a specific band portion of the echo signal for spectral analysis.

[0068] The two-dimensional scanning galvanometer, located at the beam exit of the assembly, consists of two orthogonally arranged galvanometer motor-driven reflectors on the X and Y axes, used to deflect and control the projection angle of the composite beam.

[0069] During the system installation phase, the zero-point distance of the ranging unit and the initial deflection angle of the scanning coordinate system are calibrated by driving the two-dimensional scanning galvanometer to scan the known fixed calibration point on the stator side, thereby establishing the three-dimensional rectangular coordinate system of the detection assembly itself.

[0070] S130 establishes a full-circumferential spatiotemporal synchronization reference by using an absolute position encoder installed at the main shaft end of the hydro generator as the spatiotemporal reference source of the system. The position encoder outputs a digital angle signal or pulse signal that strictly corresponds to the mechanical position of the rotor. The central processing unit acquires the signal through a high-speed IO interface and maps the continuous measurement in the time domain to the spatial angle domain of the rotor.

[0071] The set sampling resolution is Point / circle. Define the rotation angle coordinate system. The 360-degree mechanical angle of one rotation of the rotor is divided into... A discrete sampling sector. Each sampling angle position Defined as:

[0072] ;

[0073] In the formula, Indicates the first The position of each sampling angle Indicates the system's sampling resolution. The position index represents the sampling angle, and its value range is... , This indicates the complete mechanical rotation angle.

[0074] The central processing unit maintains an angle index counter. When the real-time angle value fed back by the position encoder matches... At this time, the system generates a hardware interrupt or trigger pulse, which is defined as a spatial trigger event. The spatial trigger event is simultaneously connected to the trigger input terminals of the fluorescence excitation source, the ranging laser, and the spectral acquisition circuit in the four-dimensional optical detection assembly, ensuring that the acquisition actions of all physical quantities are strictly aligned with the same rotor angle position in time, thereby eliminating the spatial position slip error caused by speed fluctuations.

[0075] S2. Perform a global inspection and simultaneously collect raw physical quantity signals; during generator operation, trigger the multimodal optical detection assembly based on the spatiotemporal synchronization reference to simultaneously collect fluorescence attenuation signals, frequency-modulated continuous wave beat frequency signals, and environmental background spectrum signals at multiple discrete sampling angle positions.

[0076] Specifically, to ensure spatial consistency of multidimensional physical quantity measurements under high-speed generator rotation conditions, step S2 includes the following sub-steps:

[0077] S210: Generate an angle synchronization trigger command. After entering the global inspection mode, the two-dimensional scanning galvanometer is maintained at a preset zero position or a fixed deflection angle, ensuring that the probe optical axis is perpendicular to the tangent plane of the rotor lead's rotation trajectory. The high-speed comparator inside the central processing unit monitors the current mechanical angle value fed back by the position encoder in real time. When the current mechanical angle value matches any sampling angle position in the angle index sequence established in step S130... When the values ​​are equal, the central processing unit immediately outputs a hardware synchronization trigger signal of TTL or LVDS level. The rising edge of this trigger signal is defined as the zero point of the current sampling period. .

[0078] S220, performing coordinated excitation and modulation of multi-band light sources, after the four-dimensional optical detection assembly receives the synchronous trigger signal, the internal light source driving circuit simultaneously performs the following actions:

[0079] A short-pulse laser is emitted by driving a fluorescent excitation source. The pulse width is usually set to the nanosecond level (e.g., 10 ns to 50 ns), and the center wavelength is matched to the excitation peak of the coating material.

[0080] The frequency-modulated continuous wave ranging light source is driven to start one linear frequency-modulated sweep cycle, generating a continuous laser beam whose frequency changes linearly with time.

[0081] The electronic shutter or integral gating of the spectral acquisition unit is activated to begin accumulating the photon energy entering the slit.

[0082] It achieves the coverage of the same rotor surface area by optical signals of different bands and different modulation modes at the same physical moment under a single trigger command.

[0083] S230, parallel acquisition of multiple channels of raw physical quantity signals, within a preset sampling time window after the trigger signal. Within, three sets of raw signals are captured in parallel through independent photoelectric detection channels, within a time window. The length is less than the time required for the rotor to rotate through one sampling sector.

[0084] The first channel acquires fluorescence attenuation signals. Photomultiplier tubes (PMTs) or avalanche photodiodes (APDs) are used to detect the fluorescent photons returning after being excited by the coating. The acquisition card records the complete decay process of light intensity over time at a high sampling rate. The signal characterizes the energy dissipation properties of the coating in the time domain.

[0085] The second channel acquires frequency-modulated continuous wave beat frequency signals. The interference signal, resulting from the mixing of the echo reflected from the lead surface of the ranging laser and the local reference light, is received using a photodiode. The signal is a sinusoidal or quasi-sinusoidal wave in the time domain, and its frequency characteristics contain information about the optical path difference. The mathematical expression of the beat frequency signal is recorded as follows:

[0086] ;

[0087] In the formula, This represents the collected continuous beat frequency signal, used as time. The function, Indicates the amplitude of the signal. Represents a sine wave function, carrier frequency signal. Indicates the frequency of the beat frequency signal. This indicates the initial phase of the beat frequency signal. This indicates system noise.

[0088] The third channel acquires ambient background spectral signals. Using a linear CCD or CMOS sensor in a miniature spectrometer, data is recorded within a preset wavelength range. The signal contains background light information other than fluorescence and ranging laser light, and is mainly composed of ambient scattered light reflected from the coating surface.

[0089] S240, construct the multidimensional raw dataset. The central processing unit packages the data collected from the above three channels and adds the current sampling angle. The tags form the first The original data tuple of each sampling point :

[0090] ;

[0091] In the formula, Indicates the first The original data tuples of each sampling point Indicates the first The angle value of each sampling point. Indicates the first A set of scalar values ​​for each sampling point Indicates the first Fluorescence intensity at each sampling point Over time Changes, Indicates the first Beat signal at each sampling point Over time Changes, Indicates the first spectral power at each sampling point With wavelength The relationship.

[0092] As the rotor continues to rotate, steps S210 to S240 are repeated until the entire circumference is completed. The acquisition of data from each point forms a raw data stream sequence covering the entire circumference of the rotor leads. Based on the mechanism of single-point angle triggering and multi-dimensional signal synchronous acquisition, the spatiotemporal misalignment between measurements of different physical quantities is eliminated from the data source. This ensures the strict correspondence between the reconstructed temperature field, morphology field and chemical fingerprint field in spatial coordinates, providing a reliable data foundation for multi-physical quantity correlation analysis.

[0093] S3. Demodulate the original physical quantity signal and reconstruct the multidimensional state field; perform time-domain demodulation on the fluorescence decay signal to calculate the fluorescence lifetime and reconstruct the one-dimensional circumferential temperature distribution field of the rotor lead; perform frequency domain analysis on the frequency-modulated continuous wave beat frequency signal to calculate the instantaneous distance and reconstruct the three-dimensional geometric profile and vibration state field of the rotor lead; perform characteristic peak integration on the environmental background spectral signal to reconstruct the environmental chemical fingerprint distribution field.

[0094] Specifically, step S300 in this embodiment includes the following sub-steps:

[0095] S310, Reconstruct a one-dimensional circumferential temperature distribution field based on time-domain fluorescence signals. The central processing unit processes each sampling angle acquired in step S200. Corresponding fluorescence decay signal Time-domain demodulation is performed. The discrete fluorescence intensity data are fitted using a single-exponential or multi-exponential decay model via the nonlinear least squares method. The fitted model is expressed as:

[0096] ;

[0097] In the formula, For a moment light intensity, The initial light intensity, Let be the optical lifetime constant to be solved. The background noise constant is... For time, It is an exponential function.

[0098] Obtain fluorescence lifetime Then, based on the temperature-life characteristic curve obtained in advance in a constant temperature oil bath or blackbody furnace, the absolute temperature value at that angular position is calculated. The temperature demodulation function is expressed using a polynomial:

[0099] ;

[0100] In the formula, For angle Temperature value at that location, For the first The fitting coefficients of order, For angle The corresponding attenuation constant, Let be the order of the polynomial. The index for summation.

[0101] The calculated temperature values ​​of all sampling points are mapped in angular order to construct a one-dimensional circumferential temperature distribution field reflecting the thermal state of the rotor leads. .

[0102] S320 reconstructs three-dimensional geometric contours and vibration state fields based on frequency domain beat frequency signals. The central processing unit processes each sampling angle. The corresponding frequency-modulated continuous wave beat frequency signal Perform a Fast Fourier Transform (FFT) to convert the time-domain signal into a frequency-domain spectrum. Use a peak search algorithm to extract the dominant frequency component from the spectrum. This frequency corresponds to the optical path difference between the measurement optical path and the reference optical path.

[0103] Based on the principle of frequency-modulated continuous wave ranging, the instantaneous distance from the detection assembly to the rotor lead surface is calculated. :

[0104] ;

[0105] In the formula, For angle instantaneous distance at the location, The speed of light in a vacuum. For the laser frequency modulation cycle, For laser modulation bandwidth, For angle The relevant beat frequency.

[0106] Combined with the rotation radius of the rotor leads and current angle Distance data in polar coordinate system Convert the data to a Cartesian coordinate system and reconstruct the spatial geometric profile of the rotor leads at the current moment.

[0107] To separate static morphology from dynamic vibration, a sliding time window is maintained for past... The static reference distance for the angular position is obtained by statistically averaging the distance data from each rotation cycle. Calculate the instantaneous distance of the current period. Distance from static reference The difference is used to obtain the radial vibration amplitude at that angular position. :

[0108] ;

[0109] In the formula, For angle Radial vibration amplitude at the location, From the current perspective Distance value at that location, This is the static reference distance at that angular position. To adjust the sliding window size, For the first Historical distance data for each period, representing that angle The distance value in the previous period, It represents a certain angular position in the system. It represents a moment in history.

[0110] S330, based on wavelength domain spectral signals, reconstructs the environmental chemical fingerprint distribution field. The central processing unit processes each sampling angle. Corresponding environmental background spectral signal Perform spectral characteristic analysis. For the target chemical component to be monitored (e.g., oil mist scattering characteristics or ozone absorption / emission characteristics), preset its characteristic response band range. .

[0111] The spectral signal is integrated within the characteristic band, and the baseline spectral intensity is subtracted to calculate the spectral characteristic intensity index at that angular position. :

[0112] ;

[0113] In the formula, To represent a certain physical quantity and Let represent the starting and ending wavelengths of the integration, respectively. Indicates at wavelength Below, the spectral power or other wavelength-related physical quantities that are specifically measured or calculated. This represents the wavelength dependence of the reference value. It represents a tiny change in wavelength.

[0114] index The relative abundance of target chemical substances along the optical path and in the layer adhering to the lead surface was quantified. The characteristic indices along the entire circumference were arranged angularly to construct the environmental chemical fingerprint distribution field. .

[0115] S4. Identify anomalies and lock the target area based on a one-dimensional circumferential temperature distribution field; compare the temperature data in the one-dimensional circumferential temperature distribution field with a preset threshold; when a temperature anomaly is detected, lock the center angle coordinates of the anomaly area according to the spatiotemporal synchronization reference, and trigger active tracking and local scanning mode.

[0116] Specifically, step S4 in this embodiment includes the following sub-steps:

[0117] S410, perform threshold discrimination analysis based on the one-dimensional circumferential temperature distribution field, and the central processing unit retrieves the reconstructed one-dimensional circumferential temperature distribution field data from step S3. It includes all the rotations within one revolution of the rotor. The absolute temperature value at each sampling angle position. The system sets a temperature alarm threshold. This threshold is determined based on the heat resistance rating of the insulation material and the statistical upper limit of historical operating data. The logic operation unit within the central processing unit traverses the entire temperature distribution sequence. Compare the temperature values ​​at the current angle point by point. Temperature alarm threshold The size relationship. The judgment logic is expressed as:

[0118] ;

[0119] In the formula, This represents the difference between the temperature and the alarm value. Indicates the first The current temperature of a given moment, location, or physical system. This indicates the set threshold for temperature alarm.

[0120] If for any Calculation results If the current rotor lead condition is normal, the system remains in global inspection mode and continues to perform full circumferential data acquisition for the next cycle. If at least one , making If a thermal anomaly is detected at that location, an abnormal interruption signal will be generated immediately.

[0121] S420, locks the spatial angular coordinates of the abnormal target area, and in response to the abnormal interruption signal, the central processing unit checks if the condition meets the requirements. Cluster analysis is performed on the data points. Due to the thermal diffusion effect, outliers typically appear as a continuous angular interval. The system employs a peak search algorithm to locate the index corresponding to the highest temperature point within the continuous over-temperature range. This point is determined as the center of the abnormal region.

[0122] Based on the spatiotemporal synchronization benchmark established in step S130, the index is retrieved. Corresponding physical space angle value And record it in the system's fault tracing register, and at the same time, Centered on the preset scanning field of view Define the target area to be scanned. .

[0123] S430 executes the logical switching and timing scheduling of the system's operating mode, and locks the center angle coordinates of the abnormal region. Subsequently, the central processing unit modifies the system's internal status register, changing the operating mode flag from global inspection to active tracking.

[0124] During this switching process, the central processing unit performs timing prediction calculations based on rotor dynamics. It reads the current instantaneous angular velocity fed back from the position encoder. and current rotor angle Calculate the predicted arrival time of the center of the anomaly region rotating back to the position aligned with the optical axis of the four-dimensional optical detection assembly. The prediction model is as follows:

[0125] ;

[0126] in, Indicates the predicted arrival time. Indicates the current moment. This represents the phase difference / angular displacement difference between the current angle and the target angle. This represents the current instantaneous angular velocity, which is obtained in real time from the position encoder.

[0127] like ,but ;

[0128] like ,but .

[0129] Based on calculations The central processing unit (CPU) reconfigures the direct memory access (DMA) channel and interrupt priority at the hardware level. The system suspends low-frequency full-circuit data transfer tasks targeting non-target areas, freeing up data bus bandwidth. Simultaneously, it preloads the control waveform data of the 2D scanning galvanometer into the cache and adjusts the sampling rate of the data acquisition card to its highest level, placing it in a ready-to-trigger state, waiting for a specified time. Upon arrival, local high-density sampling is initiated immediately.

[0130] S5. Perform distance-corrected active tracking and local high-resolution scanning; calculate the geometric correction deflection parameters of the two-dimensional scanning galvanometer based on the instantaneous distance value corresponding to the center angle coordinate in the three-dimensional geometric contour and vibration state field; drive the multi-modal optical detection assembly to perform grating scanning on the abnormal area to generate a two-dimensional high-resolution temperature matrix and a two-dimensional spectral fingerprint matrix.

[0131] Specifically, step S5 in this embodiment includes the following sub-steps:

[0132] S510, determine the temporal and spatial windows for active tracking. The central processing unit uses the predicted arrival time calculated in step S430 as a basis. Combined with the preset local scan physical width and the current linear velocity of the rotor Determine the start time of active tracking. and end time .

[0133] During the waiting time window, the value corresponding to the center angle of the abnormal region, stored in the data buffer, is invoked. Instantaneous distance value Distance value This refers to the measured vertical distance from the optical axis exit point of the probe assembly to the rotor lead surface, obtained in step S320 via the frequency-modulated continuous wave ranging unit. This measured distance is used as the geometric reference parameter for subsequent scanning control, eliminating focal plane positioning deviations caused by installation errors or rotor eccentricity.

[0134] S520 calculates the geometric correction deflection parameters based on the measured distance. In order to generate a distortion-free rectangular scanning area on the curved rotor lead surface, the central processing unit uses a geometric projection algorithm to map the coordinates of the scanning points in physical space into the angular displacement control signal of the two-dimensional scanning galvanometer.

[0135] The local scan area is set to be divided into Each high-resolution pixel. For any pixel within the scanned area. Its coordinates in the physical tangent plane coordinate system with the anomaly point as the origin are: The central processing unit determines the distance based on the measured distance. Calculate the X-axis mirror deflection angle corresponding to this pixel. and Y-axis galvanometer deflection angle The correction calculation formula is as follows:

[0136] ;

[0137] ;

[0138] In the formula, Indicates the first The lateral physical offset of a column pixel relative to the scan center. Indicates the first The vertical physical offset of a row of pixels relative to the scan center. This represents the measured distance / target distance. Represents pixels The corresponding X-axis galvanometer deflection angle, Represents pixels The corresponding Y-axis galvanometer deflection angle.

[0139] By introducing measured distance As a denominator, dynamic field scaling is achieved. When the rotor lead is far from the detector, the deflection angle is automatically reduced to maintain the physical scanning range unchanged.

[0140] When the distance is close, the deflection angle is automatically increased. This distance-sensing-assisted dynamic scanning mechanism ensures that regardless of the rotor's eccentric position, the scanned thermal and spectral maps always correspond to the actual surface of the rotor. The physical region ensures spatial consistency in imaging.

[0141] The S530 performs raster-type high-resolution scanning acquisition when the system clock reaches the start time. The central processing unit outputs the voltage sequence calculated in step S520 to the driver of the two-dimensional scanning galvanometer via a digital-to-analog converter (DAC), controlling the galvanometer reflector to deflect rapidly according to the preset grating scanning path.

[0142] While the galvanometer performs the scanning action, the four-dimensional optical detection assembly switches to high-frequency acquisition mode. The fluorescence thermography unit emits high-repetition-rate pulsed laser light to scan each pixel along the path. Perform single-point excitation and lifetime measurement to obtain the local temperature value at that point. Simultaneously, the spectral acquisition unit records the spectral intensity data at the pixel location and extracts environmental spectral feature values. .

[0143] S540 generates a multi-dimensional, high-resolution state matrix. The central processing unit indexes the time-series data stream acquired during the scanning process according to the spatial index of the scanning path. The data is reorganized to construct a two-dimensional data matrix that reflects the fine features of the abnormal region.

[0144] Generate a two-dimensional high-resolution temperature matrix Its elements Corresponding pixel The matrix provides precise temperature values ​​and visually represents the distribution of hotspots within the abnormal region.

[0145] Generate a two-dimensional spectral fingerprint matrix Its elements Corresponding pixel The spectral response intensity of a specific chemical component.

[0146] S6. Integrate multi-source feature data for intelligent fault type diagnosis; extract multi-dimensional fault feature vectors containing thermal, mechanical and chemical properties from the one-dimensional circumferential temperature distribution field, the three-dimensional geometric contour and vibration state field, the two-dimensional high-resolution temperature matrix and the two-dimensional spectral fingerprint matrix; input the multi-dimensional fault feature vectors into a preset multi-physical quantity correlation diagnosis model, and output the fault type and location result of the rotor lead;

[0147] Specifically, step S6 in this embodiment includes the following sub-steps:

[0148] S610, construct a multi-dimensional fault feature vector. The central processing unit gathers the macroscopic physical field data reconstructed in step S3 and the local microscopic matrix data generated in step S, and performs feature extraction on the locked abnormal area.

[0149] From a two-dimensional high-resolution temperature matrix Extract thermal features. Calculate the local maximum temperature value within the anomalous region. This characterizes the severity of the fault. Simultaneously, the spatial gradient of the temperature distribution is calculated by performing spatial convolution on the temperature matrix using either the Sobel operator or the Laplace operator. The spatial gradient value The concentration of hotspots was quantified: high gradient values ​​represent point heat sources, while low gradient values ​​represent diffuse heat sources.

[0150] From the three-dimensional geometric contour and vibrational field In the middle, extract the corresponding abnormal angles. radial vibration amplitude This characterizes the mechanical stability of the lead wire under the action of centrifugal force.

[0151] From two-dimensional spectral fingerprint matrix Chemical features were extracted. The integral values ​​of spectral intensity corresponding to the ozone characteristic bands were calculated respectively. and the spectral intensity integral value corresponding to the characteristic band of oil mist :

[0152] ;

[0153] In the formula, Represents a multidimensional fault feature vector. Indicates the transpose symbol. This represents the maximum temperature value among temperature-related characteristic quantities. Indicates temperature gradient characteristics, Indicates the characteristics of vibration amplitude. Indicates ozone-related intensity / concentration characteristics. This indicates the intensity or concentration characteristics of oil mist / oil gas / oil liquid.

[0154] S620 establishes a multi-physical quantity correlation diagnostic model. This model has pre-defined threshold values ​​for various characteristic parameters, including a temperature threshold. Gradient threshold Vibration threshold and spectral intensity threshold These threshold parameters are derived from the historical operating database of the generator sets and fault simulation experimental data.

[0155] Diagnostic models use logic gates or software algorithms to analyze feature vectors. The components in the data are subjected to joint Boolean operations to distinguish different fault mechanisms.

[0156] S630 performs cross-domain correlation fault identification. The central processing unit inputs the constructed feature vector into the diagnostic model and performs classification and diagnosis based on the following logical criteria:

[0157] Mechanical connection fault diagnosis. If detected... and Meanwhile, spectral characteristics and All below If the system determines the fault type as loose mechanical connection or fastener fatigue, the overheating is caused by Joule heating due to increased contact resistance, which is usually accompanied by mechanical loosening of the lead connector and increased micro-vibration.

[0158] Insulation defect fault detection. If detected... and At the same time If the system determines the fault type as partial discharge or insulation breakdown, then the fault type is either partial discharge or insulation breakdown. In this scenario, the partial arc discharge caused by insulation defects generates extremely high-temperature hotspots and simultaneously ionizes the air to produce ozone. This method achieves specific identification of electrical faults by correlating thermal morphology with chemical products.

[0159] Identifying and diagnosing heat dissipation anomalies. If detected... but At the same time If the system determines the fault type as heat dissipation obstruction due to oil contamination, then the leaked bearing oil covers the lead surface, attracting dust and forming sludge, which hinders heat dissipation and causes a large-area temperature rise. This fault is usually not accompanied by severe vibration or discharge. By correlating thermal dispersion characteristics with oil mist spectral fingerprints, this type of non-physical damage maintenance fault can be effectively distinguished.

[0160] S640 outputs comprehensive diagnostic results. Based on the above judgment results, the central processing unit generates a result including the spatial location (angle) of the fault. The system generates diagnostic reports containing fault type codes and corresponding feature vector data. These reports are sent to the host computer at the power plant's central control center via industrial Ethernet or fiber optic communication interfaces. This data guides maintenance personnel in developing targeted repair plans, achieving a closed-loop function from condition monitoring to decision support.

Claims

1. A method of fusion splicing fiber optic sensing for temperature monitoring of a hydroelectric generator rotor lead, comprising: Includes the following steps: S1. Construct a multimodal optical sensing environment and establish a spatiotemporal synchronization reference; Deploy a multimodal optical detection assembly integrating a fluorescence temperature measurement unit, a frequency-modulated continuous wave ranging unit, a spectrum acquisition unit and a two-dimensional scanning galvanometer on the generator stator side; Prepare a fluorescence-matrix composite functional coating on the rotor lead surface; Establish a full-circumferential spatiotemporal synchronization reference using a position encoder. S2. Based on the spatiotemporal synchronization reference, the multimodal optical detection assembly is triggered to perform a global inspection, and fluorescence attenuation signal, frequency-modulated continuous wave beat frequency signal and environmental background spectrum signal are simultaneously collected at multiple discrete sampling angle positions. S3. Demodulate the original physical quantity signal and reconstruct the multidimensional state field; demodulate the fluorescence attenuation signal in the time domain to reconstruct the one-dimensional circumferential temperature distribution field; analyze the frequency domain of the frequency-modulated continuous wave beat signal to reconstruct the three-dimensional geometric contour and vibration state field; integrate the characteristic peaks of the environmental background spectral signal to reconstruct the environmental chemical fingerprint distribution field. S4. Compare the one-dimensional circumferential temperature distribution field with a preset threshold. When an anomaly is detected, lock the center angle coordinates of the abnormal area according to the spatiotemporal synchronization reference and trigger active tracking and local scanning mode. S5. Calculate the geometric correction deflection parameters of the two-dimensional scanning galvanometer based on the instantaneous distance value corresponding to the center angle coordinate in the three-dimensional geometric contour and vibration state field, and drive the multi-modal optical detection assembly to perform grating scanning on the abnormal area to generate a two-dimensional high-resolution temperature matrix and a two-dimensional spectral fingerprint matrix. The calculation of the geometric correction deflection parameters of the two-dimensional scanning galvanometer in step S5 is specifically as follows: The measured vertical distance corresponding to the center angle coordinates, stored in the three-dimensional geometric contour and vibration state field, is called as the reference distance parameter; For any pixel within a local scanning area, obtain its horizontal and vertical physical offsets relative to the scanning center. Calculate the ratio of the lateral physical offset to the measured vertical distance, and take the arctangent of this ratio as the X-axis galvanometer deflection angle; Calculate the ratio of the longitudinal physical offset to the measured vertical distance, and take the arctangent of this ratio as the Y-axis galvanometer deflection angle; The voltage sequence generated by the X-axis galvanometer deflection angle and the Y-axis galvanometer deflection angle drives the two-dimensional scanning galvanometer to eliminate imaging distortion caused by changes in rotor distance; S6. Extract multidimensional fault feature vectors containing thermal, mechanical and chemical properties from the one-dimensional circumferential temperature distribution field, the three-dimensional geometric contour and vibration state field, the two-dimensional high-resolution temperature matrix and the two-dimensional spectral fingerprint matrix; input the multidimensional fault feature vectors into a preset multi-physical quantity correlation diagnostic model, and output the fault type and location result of the rotor lead.

2. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 1, characterized in that, In step S1, the fluorescent-matrix composite functional coating is prepared by mixing rare earth ion-doped inorganic phosphor with a high reflectivity ceramic matrix material and then using an atmospheric plasma spraying process. The high reflectivity ceramic matrix material is alumina or zirconium oxide. The multimodal optical detection assembly adopts a common optical path design, wherein the two-dimensional scanning galvanometer is located at the beam exit port of the multimodal optical detection assembly, and is used to uniformly deflect and control the detection beams of the fluorescence thermometry unit, the frequency-modulated continuous wave ranging unit and the spectral acquisition unit.

3. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 1, characterized in that, Step S2, which triggers the multimodal optical detection assembly based on the spatiotemporal synchronization reference, includes: The mechanical angle of one revolution of the rotor is divided into multiple discrete sampling sectors, and a unique angle index is assigned to each sampling sector. A spatial trigger event is generated when the real-time angle value fed back by the position encoder matches any sampled angle position; In response to the spatial triggering event, the fluorescence thermometry unit is simultaneously driven to emit a pulsed laser, the frequency-modulated continuous wave ranging unit is driven to start a linear frequency-modulated sweep cycle, and the electronic shutter of the spectral acquisition unit is opened, so as to acquire the fluorescence attenuation signal, the frequency-modulated continuous wave beat frequency signal, and the environmental background spectral signal in parallel within the same time window.

4. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 1, characterized in that, The demodulation of the original physical quantity signal and the reconstruction of the multidimensional state field in step S3 specifically includes: The fluorescence decay signal is fitted with an exponential decay model using the nonlinear least squares method to obtain the fluorescence lifetime constant, and the absolute temperature value is calculated based on the calibrated temperature-lifetime characteristic curve to construct the one-dimensional circumferential temperature distribution field. Perform a fast Fourier transform on the frequency-modulated continuous wave beat frequency signal, extract the main frequency component to calculate the probe optical path, and combine the rotor rotation radius to convert the distance data in the polar coordinate system into the rectangular coordinate system data to construct the three-dimensional geometric profile and vibration state field; For the target chemical component to be monitored, a characteristic response band range is preset. The environmental background spectral signal is integrated within the characteristic response band range and the baseline intensity is subtracted to calculate the spectral characteristic intensity index and construct the environmental chemical fingerprint distribution field.

5. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 1, characterized in that, Step S4 involves locking the center angle coordinates of the abnormal region and triggering active tracking and local scanning modes, including: Calculate the difference between the temperature data in the one-dimensional circumferential temperature distribution field and the preset temperature alarm threshold. When the difference is greater than zero, generate an abnormal interruption signal. Peak search is performed on the continuous over-temperature range to determine the center angle coordinates of the abnormal region; Based on the current instantaneous angular velocity and current rotor angle fed back by the position encoder, the predicted arrival time of the center angle coordinate rotating to the optical axis alignment position of the multimodal optical detection assembly is calculated; Based on the predicted arrival time, the scheduling system enters active tracking mode and preloads the control data of the two-dimensional scanning galvanometer.

6. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 1, characterized in that, Step S5, which generates a two-dimensional high-resolution temperature matrix and a two-dimensional spectral fingerprint matrix, includes: Within the active tracking time window, the two-dimensional scanning galvanometer is controlled to deflect according to a preset grating path; During the scanning process, the fluorescence thermometry unit is used to obtain the local temperature values ​​of each pixel on the scanning path, forming the two-dimensional high-resolution temperature matrix. Simultaneously, the environmental spectral feature values ​​of each pixel on the scanning path are acquired using the spectral acquisition unit to form the two-dimensional spectral fingerprint matrix; The two-dimensional high-resolution temperature matrix and the two-dimensional spectral fingerprint matrix are strictly corresponding in terms of spatial pixel index.

7. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 1, characterized in that, The process of constructing the multidimensional fault feature vector in step S6 includes: Extract the local maximum temperature value from the two-dimensional high-resolution temperature matrix and calculate the spatial gradient value of the temperature distribution; Extract the radial vibration amplitude corresponding to the center angle coordinate from the three-dimensional geometric contour and vibration state field; Extract the spectral intensity integral values ​​corresponding to the ozone characteristic band and the spectral intensity integral values ​​corresponding to the oil mist characteristic band from the two-dimensional spectral fingerprint matrix. The local maximum temperature value, the spatial gradient value, the radial vibration amplitude, the spectral intensity integral value of the ozone characteristic band, and the spectral intensity integral value of the oil mist characteristic band are normalized and combined to form the multidimensional fault feature vector.

8. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 7, characterized in that, In step S6, the multidimensional fault feature vector is input into a preset multi-physical quantity correlation diagnostic model, and the fault type of the rotor lead is output, including the following logical criteria: When the local maximum temperature value is higher than the temperature threshold, and the radial vibration amplitude is higher than the vibration threshold, while the spectral intensity integral value of the ozone characteristic band and the spectral intensity integral value of the oil mist characteristic band are both lower than their respective spectral intensity thresholds, the fault type is determined to be loose mechanical connection or fastener fatigue. When the local maximum temperature value is higher than the temperature threshold, the spatial gradient value is higher than the gradient threshold, and the spectral intensity integral value of the ozone characteristic band is higher than the spectral intensity threshold, the fault type is determined to be partial discharge or insulation layer breakdown. When the local maximum temperature value is higher than the temperature threshold, but the spatial gradient value is lower than the gradient threshold, and the spectral intensity integral value of the oil mist characteristic band is higher than the spectral intensity threshold, the fault type is determined to be heat dissipation obstruction caused by oil stains.

9. The method for monitoring the rotor lead temperature of a hydro-generator integrating fiber optic sensing according to claim 7, characterized in that, The method for calculating the radial vibration amplitude is as follows: Maintain a sliding time window and statistically average the distance data within a preset number of rotation cycles in the past to obtain the static reference distance at the center angle coordinates; Calculate the difference between the instantaneous distance of the current cycle and the static reference distance, and determine the absolute value of the difference as the radial vibration amplitude; The instantaneous distance is calculated by the frequency-modulated continuous wave ranging unit according to the linear frequency modulation ranging principle. Specifically, it is the value obtained by dividing the product of the vacuum speed of light, the laser frequency modulation period, and the main frequency component of the beat frequency signal by twice the laser modulation bandwidth.

Images