Method, device and system for diagnosing wear of guide vane seal of water pump turbine

Abstract

This application provides a method, apparatus, and system for diagnosing guide vane seal wear in water pumps and turbines. The method includes: acquiring multi-source data of guide vanes in water pumps and turbines and performing time-series integration; using the guide vane full-closure action trigger timestamp in the multi-source data as a benchmark, extracting event data blocks from the time-series integrated multi-source data, and then performing noise removal and feature extraction to obtain structured feature data; employing a two-level algorithm of guide vane-level coarse positioning and sealing surface partition fine positioning to calculate positioning features on the structured feature data, obtaining wear positioning data including faulty guide vanes and seal types; based on hydraulic parameters, structured features, and wear positioning data, determining the dominant wear type and assessing the comprehensive fault confidence, and outputting graded diagnostic results. This application can achieve accurate positioning of guide vane seal wear in scenarios with multiple guide vanes and strong hydraulic noise coupling in hydraulic machinery.

Description

Technical Field

[0001] This application relates to the field of fault diagnosis technology for hydropower equipment, and in particular to a method, device and system for diagnosing wear of guide vane seals of water pumps and turbines. Background Technology

[0002] Pump turbines are the core equipment of pumped storage power stations. Guide vanes, as key components for flow regulation and shutdown sealing, directly affect the unit's operating efficiency and safety reliability. During long-term operation, the guide vane end faces and vertical seals are subjected to high-speed water flow erosion, sediment abrasion, cavitation, and mechanical vibration, leading to gradual wear of the sealing surfaces and the generation of gap leaks.

[0003] Early wear of the guide vane seal results in minimal leakage, which traditional flow meters cannot accurately measure. Furthermore, the leakage rate is highly susceptible to variations in head, tailwater level, guide vane clearance, and unit operating conditions, making it difficult to distinguish between normal and abnormal leaks. By the time maintenance personnel detect a significant increase in leakage or the unit's inability to shut down normally, deep grooves have already formed on the sealing surface, necessitating a shutdown for replacement. Unplanned downtime losses for a single 300MW unit can exceed one million yuan.

[0004] Currently, monitoring of guide vane seal wear mainly relies on manual inspections, judging by listening to sounds and observing leakage, depending entirely on the experience of maintenance personnel, resulting in extremely high rates of misjudgment and missed detection. While some existing technologies employ vibration or acoustic emission monitoring systems, none have solved the problem of accurate positioning in hydraulic machinery with multiple guide vanes and strong hydraulic noise coupling. It has long been believed in the field that steady-state operating data, with stable operating conditions and a single noise source, is more suitable for fault diagnosis; while transient processes (such as guide vane adjustment and operating condition switching) have complex characteristics, many influencing factors, and are difficult to diagnose. Therefore, existing monitoring systems generally rely on steady-state operating data for analysis. This technological bias leads to the systematic neglect of transient strong friction signals from guide vane seal wear—in fact, the friction and impact signals generated by the sealing surface pressing together when the guide vane is fully closed have a much higher signal-to-noise ratio than the leakage turbulence signals during steady-state operation, making them a key source of information for identifying early, subtle wear. Furthermore, existing positioning schemes can only locate wear down to the single guide vane level, failing to distinguish between end face and vertical surface seal wear, requiring disassembly and inspection of the entire sealing structure during maintenance, resulting in limited efficiency improvements. Summary of the Invention

[0005] The purpose of this application is to provide a method, device and system for diagnosing guide vane seal wear in water pumps and turbines, which can achieve accurate positioning of guide vane seal wear in scenarios with multiple guide vanes and strong hydraulic noise coupling in hydraulic machinery.

[0006] Firstly, this application provides a method for diagnosing guide vane seal wear in a water pump turbine. The method includes: acquiring multi-source data of the guide vane of the water pump turbine and performing time-series integration; the multi-source data includes: guide vane action signal, acoustic emission signal, vibration signal, and hydraulic parameters, wherein the guide vane action signal includes a guide vane full-closure action trigger time stamp; based on the guide vane full-closure action trigger time stamp, extracting event data blocks of a specified duration from the time-series integrated multi-source data, removing coupling interference caused by head, tailwater level, and guide vane opening in the event data blocks through a pre-constructed hydraulic benchmark health model, and then performing feature construction to obtain structured feature data; employing a two-level algorithm of guide vane-level coarse positioning and sealing surface partition fine positioning to calculate positioning features on the structured feature data, obtaining wear positioning data including faulty guide vanes and sealing types; based on hydraulic parameters, structured features, and wear positioning data, determining the dominant wear type and evaluating the comprehensive fault confidence, and outputting graded diagnostic results.

[0007] Furthermore, the steps described above, which extract event data blocks of a specified duration from the time-series integrated multi-source data based on the guide vane full closure trigger time stamp, remove coupling interference caused by head, tailwater level, and guide vane opening in the event data blocks using a pre-constructed hydraulic benchmark health model, and then construct structured feature data, include: extracting data segments within a specified time period based on the guide vane full closure trigger time stamp from the time-series integrated multi-source data to form a single guide vane full closure event data block; removing interference signals from each guide vane full closure event data block using a pre-constructed hydraulic benchmark health model under normal operating conditions; extracting key features from the guide vane full closure event data blocks after removing interference signals to obtain structured feature data; the structured feature data includes at least: the normalized acoustic emission peak value and corresponding peak time of each guide vane, the effective number of acoustic emission pulses of each guide vane, and the normalized total vibration energy of each guide vane.

[0008] Furthermore, the aforementioned hydraulic baseline health model includes an acoustic emission baseline health model and a vibration baseline health model, and the construction process is as follows: acquire historical fault-free operation data; divide the head into multiple intervals according to the operating range; within each interval, use a quadratic polynomial to fit the relationship between the acoustic emission signal and the guide vane opening under normal operating conditions to obtain the acoustic emission baseline health model; use a quadratic polynomial to fit the relationship between the vibration signal and the guide vane opening under normal operating conditions to obtain the vibration baseline health model; or, use a deep learning algorithm to train the model based on historical fault-free operation data to construct the acoustic emission baseline health model and the vibration baseline health model.

[0009] Furthermore, the aforementioned acoustic emission reference health model includes an acoustic emission health reference signal; the vibration reference health model includes a vibration health reference signal; the step of removing interference signals from each guide vane fully closed event data block using a pre-constructed hydraulic reference health model under normal operating conditions includes: for each guide vane fully closed event data block, calculating the difference between the acoustic emission signal and the acoustic emission health reference signal in the guide vane fully closed event data block to obtain a normalized acoustic emission signal; calculating the difference between the vibration signal and the vibration health reference signal in the guide vane fully closed event data block to obtain a normalized vibration signal; and performing noise reduction processing on the normalized acoustic emission signal and the normalized vibration signal respectively based on the wavelet threshold denoising method.

[0010] Furthermore, the above-mentioned two-level algorithm of guide vane-level coarse positioning and sealing surface partition fine positioning is used to calculate the positioning features of the structured feature data to obtain wear positioning data containing faulty guide vanes and sealing types. This includes: acquiring standard data for each guide vane end face / vertical face; the standard data includes: standard propagation time of end face sealing and standard propagation time of vertical face sealing; calculating target feature data based on the structured feature data; the target feature data includes: peak time difference matrix, amplitude ratio matrix, acoustic emission anomaly coefficient of each guide vane, vibration anomaly coefficient of each guide vane, signal arrival time difference of each guide vane, signal rise time of each guide vane, and signal dominant frequency of each guide vane; and using the guide vane-level coarse positioning and sealing surface partition fine positioning two-level algorithm to calculate the positioning features based on the standard data of each guide vane end face / vertical face and the target feature data to obtain wear positioning data containing faulty guide vanes and sealing types.

[0011] Furthermore, the steps for calculating target feature data based on structured feature data include: determining the amplitude ratio matrix based on the normalized acoustic emission peak value of each guide vane; determining the peak time difference matrix based on the peak time corresponding to the normalized acoustic emission peak value of each guide vane; determining the acoustic emission anomaly coefficient based on the number of effective acoustic emission pulses of each guide vane; and determining the vibration anomaly coefficient based on the normalized total vibration energy of each guide vane.

[0012] Furthermore, the steps described above, based on the standard data and target feature data of each guide vane end / facade, employing a two-level algorithm of guide vane-level coarse positioning and sealing surface zoning fine positioning to calculate positioning features and obtain wear positioning data including faulty guide vanes and sealing types, include: traversing the normalized acoustic emission peak values ​​of all guide vanes to determine the maximum and second-largest normalized acoustic emission peak values; calculating the ratio of the maximum to the second-largest normalized acoustic emission peak values ​​to obtain the amplitude ratio; if the amplitude ratio exceeds a preset threshold, directly determining the guide vane corresponding to the maximum normalized acoustic emission peak value as a faulty guide vane; if the amplitude ratio does not exceed the preset threshold, combining... The maximum values ​​of the vibration anomaly coefficients of each guide vane are cross-validated; the actual propagation time of the faulty guide vane is calculated; the end-face sealing time difference is calculated based on the actual propagation time and the standard propagation time of the end-face seal; the vertical sealing time difference is calculated based on the actual propagation time and the standard propagation time of the vertical seal; the guide vane-level reliability is determined based on the amplitude ratio; the sealing surface reliability is determined based on the difference between the end-face sealing time difference and the vertical sealing time difference; the guide vane-level reliability and the sealing surface reliability are weighted and summed to obtain the location reliability; when the location reliability exceeds the corresponding threshold, it is judged as a high-confidence location; otherwise, it is judged as a low-confidence location and marked as an anomaly to be observed.

[0013] Furthermore, the steps described above for determining the dominant wear type and assessing the comprehensive fault confidence based on hydraulic parameters, structured features, and wear location data, and outputting graded diagnostic results, include: when an acoustic emission signal appears within a specified time after the guide vane fully closes and the acoustic emission anomaly coefficient exceeds the corresponding threshold, it is determined to be dominated by seal wear; when the head fluctuation exceeds the first preset range or the tailwater level fluctuation exceeds the second preset range and the acoustic emission anomaly coefficient is less than the corresponding threshold, it is determined to be dominated by hydraulic disturbance; the reliability of the location result is calculated based on the location reliability; and the normalized acoustic emission peak value of each guide vane and the preset moderate wear acoustic emission peak value threshold are used to determine the dominant wear type. The system calculates the wear intensity based on the acoustic emission anomaly coefficient and vibration anomaly coefficient; it calculates the signal anomaly degree based on the normalized acoustic emission peak value in the most recent specified events; it performs a weighted summation of the location result confidence, wear intensity, signal anomaly degree, and wear development trend to obtain the overall fault confidence; when the overall fault confidence exceeds the first threshold and is determined to be dominated by seal wear, it outputs a high-confidence guide vane seal wear diagnosis result; when the overall fault confidence is between the first and second thresholds and is determined to be dominated by seal wear, it outputs a medium-confidence guide vane seal wear diagnosis result; the second threshold is less than the first threshold.

[0014] Secondly, this application also provides a device for diagnosing guide vane seal wear in a water pump turbine. The device includes: a data acquisition module for acquiring multi-source data of the guide vane of the water pump turbine and performing time-series integration; the multi-source data includes: guide vane action signal, acoustic emission signal, vibration signal, and hydraulic parameters, and the guide vane action signal includes the guide vane full-closure action trigger timestamp; a feature extraction module for extracting event data blocks of a specified duration from the time-series integrated multi-source data based on the guide vane full-closure action trigger timestamp, removing coupling interference caused by head, tailwater level, and guide vane opening in the event data blocks through a pre-constructed hydraulic benchmark health model, and then constructing features to obtain structured feature data; a wear location module for performing location feature calculation on the structured feature data using a two-level algorithm of guide vane-level coarse location and sealing surface partition fine location to obtain wear location data including faulty guide vane and sealing type; and a diagnosis result output module for determining the dominant wear type and evaluating the comprehensive fault confidence based on hydraulic parameters, structured features, and wear location data, and outputting graded diagnosis results.

[0015] Thirdly, this application also provides a pump-turbine guide vane seal wear diagnosis system, the system comprising: a server and a data acquisition system connected in communication; the data acquisition system includes multi-source sensors and a power plant computer monitoring system; the data acquisition system is used to collect multi-source data of the pump-turbine guide vane, and the server is used to execute the method described in the first aspect.

[0016] The method, device, and system for diagnosing guide vane seal wear in this application for water pumps and turbines abandon the traditional mature approach of using steady-state operating data for fault diagnosis. It overcomes the technical bias of numerous transient influencing factors and difficulty in diagnosis, focusing on capturing key data before and after the guide vane fully closes. At this moment, the friction and impact signals generated during the instant the sealing surface is pressed together are strongest, allowing for the identification of earlier, subtle wear. By constructing a hydraulic baseline health model, interference from head, tailwater level, and guide vane opening is eliminated, separating early, subtle wear signals from strong background noise. A positioning algorithm based on acoustic propagation theory is introduced, which, through the time difference and amplitude ratio of acoustic emission signals between different sensors, can accurately locate the specific guide vane experiencing wear, and even pinpoint the exact location of the sealing surface, significantly improving maintenance efficiency. Employing a multi-physical quantity cross-validation and confidence quantification mechanism, comprehensively considering positioning reliability, wear intensity, signal anomaly degree, and development trend, the false alarm rate can be significantly reduced, allowing for early detection of guide vane seal wear and providing sufficient time for planned maintenance. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for diagnosing wear of guide vane seals in a water pump turbine, provided in this application embodiment; Figure 2 A structural block diagram of a water pump turbine guide vane seal wear diagnosis system provided in this application embodiment; Figure 3 This is a structural block diagram of a water pump turbine guide vane seal wear diagnostic device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of this application will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] Currently, vibration or acoustic emission technology is commonly used for diagnosing pump-turbine guide vane seals. However, these methods generally suffer from the following drawbacks: First, they fail to consider the elimination of interference from other operating factors, making it difficult to effectively eliminate noise interference caused by head fluctuations, tailrace fluctuations, and hydraulic disturbances, resulting in a high false alarm rate. Second, they cannot pinpoint which specific guide vane is worn, requiring shutdown and individual inspection after an anomaly is detected, leading to low maintenance efficiency. Third, they cannot distinguish between early weak fault signals and random noise, making early warning difficult. While there have been attempts to apply acoustic emission technology to mechanical seal monitoring, these primarily target single-sensor threshold alarms, failing to address the problem of accurate positioning under the coupling of multiple guide vanes and strong hydraulic noise in hydraulic machinery. Furthermore, interference in pump-turbine guide vane seals is characterized by strong transient nature, wide bandwidth, and strong coupling with hydraulic parameters, making the elimination of interference from other operating factors complex. Traditional methods focus on fault diagnosis under steady-state operating conditions, but the transient process is complex, has many influencing factors, and is difficult to diagnose, resulting in a long-standing technical problem of "emphasizing steady-state analysis while neglecting transient diagnosis."

[0021] Based on this, embodiments of this application provide a method, device, and system for diagnosing wear on guide vane seals of a water pump turbine. This method can accurately identify early, minor wear, precisely locate faulty guide vanes and seal types, significantly reduce false alarms, enable early warning and condition-based maintenance, and improve unit operational reliability and maintenance efficiency. To facilitate understanding of this embodiment, a detailed description of the method for diagnosing wear on guide vane seals of a water pump turbine disclosed in this application embodiment will be provided first. Figure 1 A flowchart of a method for diagnosing wear of guide vane seals in a water pump turbine, provided in this application embodiment, is shown. This method is applied to a server; the method specifically includes the following steps: Step S102: Acquire multi-source data of the water pump turbine guide vane and perform time-series integration; the multi-source data includes: guide vane action signal, acoustic emission signal, vibration signal and hydraulic parameters, and the guide vane action signal includes the guide vane full closure action trigger timestamp; The server collects multi-source data on the pump turbine guide vanes through a data acquisition system. This system includes multi-source sensors and a power plant computer monitoring system. Acoustic emission signals are acquired through acoustic emission sensors; vibration signals are acquired through vibration sensors; and pressure signals, a type of hydraulic parameter, are acquired through pressure sensors. The power plant computer monitoring system collects the opening signal of each guide vane, the turbine head, tailrace level, and the timestamp triggering the guide vane's full closure action. The turbine head and tailrace level are hydraulic parameters; the opening signal of each guide vane and the timestamp triggering the guide vane's full closure action are guide vane action signals. Timing integration aligns the timestamps of all signals.

[0022] Step S104: Based on the trigger time stamp of the guide vane fully closing action, extract event data blocks of a specified duration from the multi-source data after time-series integration. Remove the coupling interference caused by head, tail water level and guide vane opening in the event data blocks through the pre-built hydraulic benchmark health model, and then perform feature construction to obtain structured feature data. This step involves extracting data segments within a specified time period, based on the trigger time stamp of the guide vane full closure action, from the time-series integrated multi-source data to form a single guide vane full closure event data block. Then, using a pre-constructed hydraulic baseline health model under normal operating conditions, interference signals are removed from each guide vane full closure event data block. This hydraulic baseline health model is determined based on historical fault-free operation data. Finally, key features are extracted from the guide vane full closure event data blocks after removing interference signals to obtain structured feature data.

[0023] The structured feature data mentioned above includes at least: the normalized acoustic emission peak value and corresponding peak time of each guide vane, the effective number of acoustic emission pulses of each guide vane, and the normalized total vibration energy of each guide vane; it may also include: the average head, average tailwater level and average guide vane opening, volute pressure fluctuation amplitude and tailwater pressure fluctuation amplitude within the event data block.

[0024] The specific methods for event extraction, interference removal, and feature construction will be detailed later.

[0025] Step S106: Using a two-level algorithm of coarse positioning at the guide vane level and fine positioning at the sealing surface partition, the structured feature data is used to calculate the positioning features to obtain wear positioning data that includes the faulty guide vane and the sealing type. When implementing this step, standard data for each guide vane end face / facade can be acquired first. The standard data includes: standard propagation time for end face sealing and standard propagation time for facade sealing. Then, based on the structured feature data, target feature data is calculated. The target feature data includes: peak time difference matrix, amplitude ratio matrix, acoustic emission anomaly coefficient for each guide vane, vibration anomaly coefficient for each guide vane, signal arrival time difference for each guide vane, signal rise time for each guide vane, and signal dominant frequency for each guide vane. Finally, based on the standard data for each guide vane end face / facade and the target feature data, a two-level algorithm of guide vane-level coarse positioning and sealing surface partition fine positioning is used to calculate the positioning features, obtaining wear positioning data that includes faulty guide vanes and sealing types.

[0026] By employing a two-stage algorithm—coarse positioning at the guide vane level and fine positioning at the sealing surface zoning level—wearing guide vanes can be accurately determined, and even the specific location of the sealing surface can be pinpointed, significantly improving maintenance efficiency. The detailed process will be described later.

[0027] Step S108: Based on hydraulic parameters, structured features, and wear location data, determine the dominant wear type and assess the comprehensive fault confidence, and output the graded diagnosis results.

[0028] The confidence assessment methods for specific wear type analysis will be described in detail later.

[0029] The pump turbine guide vane seal wear diagnosis method provided in this application constructs a hydraulic baseline health model to eliminate interference from head, tailwater level, and guide vane opening, separating early, weak wear signals from strong background noise. It introduces a positioning algorithm based on acoustic propagation theory, consisting of a two-stage algorithm: coarse positioning at the guide vane level and fine positioning at the sealing surface level. By using the time difference and amplitude ratio of acoustic emission signals between different sensors, it can accurately locate the specific guide vane experiencing wear, and even pinpoint the exact location of the sealing surface, significantly improving maintenance efficiency. Furthermore, by employing multi-physical quantity cross-validation and confidence quantification mechanisms, it comprehensively considers positioning reliability, wear intensity, signal anomaly degree, and development trend, significantly reducing the false alarm rate and enabling early detection of guide vane seal wear, providing sufficient time for planned maintenance.

[0030] This application also provides a water pump turbine guide vane seal wear diagnosis system. See [link to relevant documentation] Figure 2As shown, the pump turbine guide vane seal wear diagnosis system provided in this application embodiment includes a data acquisition module 100, a preprocessing module 200, a feature positioning module 300, and a diagnosis module 400; 1. Data acquisition module 100 The data acquisition module 100 includes a multi-source acquisition unit 101 and a timing integration unit 102.

[0031] The multi-source acquisition unit 101 is used to deploy multi-source sensors and combine them with the power plant computer monitoring system to acquire guide vane action signals, acoustic emission signals, vibration signals and hydraulic parameters.

[0032] (1) Acoustic emission sensor: One piezoelectric acoustic emission sensor, model R15a, with a resonant frequency of 150kHz and a sampling frequency of 1MHz, is deployed on the outer wall of the sleeve of each guide vane. It is rigidly connected to the guide vane body to eliminate the propagation path of mixed media such as oil film and air gap. For a unit with N guide vanes, a total of N acoustic emission sensors are deployed, denoted as AE1~AEN, corresponding to guide vanes No. 1 to No. N respectively.

[0033] (2) Vibration sensors: One acceleration vibration sensor, model YD-12, with a sensitivity of 100mV / g and a sampling frequency of 10kHz, is deployed in the middle of each guide vane arm. A total of N vibration sensors are deployed, denoted as a1~aN.

[0034] (3) Pressure sensor: A high-precision pressure sensor P1 is deployed at the inlet flange of the volute and a high-precision pressure sensor P2 is deployed at the outlet flange of the tailwater pipe. The sampling frequency of both is 1kHz.

[0035] (4) Computer monitoring system interface: Connect to the power plant computer monitoring system via Modbus protocol to obtain the opening signal of each guide vane in real time. Unit head Tailwater level and guide vane full closure action trigger timestamp .

[0036] (5) Sealing calibration: Standard lead-breaking tests were performed at the end face seal and vertical seal positions of each guide vane, and the sound velocity of the end face seal calibration was measured and recorded. Facade sealing calibration sound velocity Standard propagation time of end face Facade Standards Dissemination Time The calibration cycle is quarterly or after a major overhaul of the unit.

[0037] The timing integration unit 102 is used to align the timestamps of all signals, employing GPS timing to ensure that the time synchronization error of all sensors and the computer monitoring system is less than 1ms. After alignment, a running dataset S1 is formed, with each data entry containing the following fields: timestamp t, ... ~ , ~ , , , ~ , , Action marker. The action marker takes the value 0 or 1, where 1 indicates that the timestamp t is the trigger time for the guide vane full closure action.

[0038] 2. Preprocessing module 200 The preprocessing module 200 includes an event extraction unit 201, an interference removal unit 202, and a feature construction unit 203.

[0039] The event extraction unit 201 is used to extract data segments within a specified time period based on the trigger timestamp of the guide vane full closure action from the multi-source data after time-series integration, forming a single guide vane full closure event data block; In practice, the trigger time stamp is the full closure action of the guide vane. Based on this, a data segment of 20 seconds, consisting of the first 5 seconds and the last 15 seconds, is extracted to form a single guide vane fully closed event data block B, meaning each event data block contains 20 seconds of data. The reason for choosing the guide vane fully closed action as the event trigger point is that the guide vane sealing surface generates strong compression and friction at the moment of full closure, at which point the acoustic emission and vibration signals generated by wear are strongest, resulting in the highest signal-to-noise ratio.

[0040] Interference removal unit 202 is used to remove interference signals in each guide vane full closure event data block by using a pre-built hydraulic reference health model under normal operating conditions. In practice, a hydraulic baseline health model under normal operating conditions is established to eliminate interference from head, tailwater level, and guide vane opening on the signal. This hydraulic baseline health model includes an acoustic emission baseline health model and a vibration baseline health model. The construction process is as follows: Historical fault-free operation data is obtained; the head is divided into multiple intervals according to the operating range; within each interval, a quadratic polynomial is used to fit the relationship between the acoustic emission signal and the guide vane opening under normal operating conditions to obtain the acoustic emission baseline health model; a quadratic polynomial is used to fit the relationship between the vibration signal and the guide vane opening under normal operating conditions to obtain the vibration baseline health model; alternatively, a deep learning algorithm is used to train the model based on historical fault-free operation data to construct the acoustic emission baseline health model and the vibration baseline health model.

[0041] In practice: (1) Collect the fault-free operation data of the unit in the past 3 months, and screen out the guide vane full closure event data block, with a total of no less than 100 samples.

[0042] (2) The water head H is divided into multiple intervals according to the operating range. In this embodiment, it is divided into 5 intervals: H<500m, 500m≤H<520m, 520m≤H<540m, 540m≤H<560m, and H≥560m.

[0043] (3) Within each head interval, a quadratic polynomial is used to fit the relationship between the acoustic emission signal and the guide vane opening to obtain the acoustic emission baseline health model: Where a0, a1, and a2 are fitting coefficients, solved using the least squares method. As a preferred embodiment, a baseline acoustic emission health model can also be constructed based on historical normal data using algorithms such as deep learning. This baseline health model is obtained by dividing the historical normal data into training and testing sets. Of course, considering the long-term normal aging of the unit and changes in equipment characteristics after maintenance, the acoustic emission baseline health model needs to be updated periodically or after maintenance.

[0044] (4) Using the same method, the relationship between the vibration signal and the guide vane opening was fitted to obtain the vibration baseline health model. As a preferred embodiment, a vibration baseline health model can also be constructed based on historical normal data using algorithms such as deep learning. This baseline health model is obtained by dividing the historical normal data into training and testing sets. Of course, considering the long-term normal aging of the unit and changes in equipment characteristics after maintenance, the vibration baseline health model needs to be updated periodically or after maintenance.

[0045] The aforementioned acoustic emission reference health model includes an acoustic emission health reference signal; the vibration reference health model includes a vibration health reference signal; the step of removing interference signals from each guide vane fully closed event data block using a pre-constructed hydraulic reference health model under normal operating conditions includes: for each guide vane fully closed event data block, calculating the difference between the acoustic emission signal and the acoustic emission health reference signal in the guide vane fully closed event data block to obtain a normalized acoustic emission signal; calculating the difference between the vibration signal and the vibration health reference signal in the guide vane fully closed event data block to obtain a normalized vibration signal; and performing noise reduction processing on the normalized acoustic emission signal and the normalized vibration signal respectively based on the wavelet threshold denoising method.

[0046] In practice, for each new event data block, the normalized acoustic emission signal is calculated: , This represents the normalized acoustic emission signal corresponding to the i-th guide vane at time t; This represents the acoustic emission signal corresponding to the i-th guide vane at time t; Represent the acoustic emission health reference signal corresponding to the i-th guide vane; and calculate the normalized vibration signal: ; This represents the normalized vibration signal corresponding to the i-th guide vane at time t; This represents the vibration signal of the i-th guide vane at time t; Let represent the vibration health baseline signal corresponding to the i-th guide vane at time t. Then, the normalized acoustic emission signal and vibration signal are denoised using a wavelet thresholding method. The wavelet basis is db5, the decomposition level is 5, the threshold is determined by a logarithmic thresholding strategy, and a soft thresholding function is used for thresholding.

[0047] During long-term operation of the unit, the seals themselves will undergo normal aging, and the baseline health model should be updated with new data every six months or year. In addition, the baseline health model also needs to be rebuilt after a major overhaul.

[0048] Feature construction unit 203 is used to extract key features from the guide vane fully closed event data block after removing interference signals, obtaining structured feature data and forming structured record S2. The extracted features include: (1) Normalized acoustic emission peak value of each guide vane and corresponding peak time ; (2) Number of effective acoustic emission pulses per guide vane That is, the number of pulses with amplitudes exceeding three times the noise level; (3) Normalized total vibrational energy of each guide vane ; (4) Average head within the event data block Average tailwater level and average guide vane opening ; (5) Volute pressure fluctuation amplitude and tailwater pressure fluctuation amplitude .

[0049] Each data entry in structured record S2 includes: event ID, guide leaf number, , , , , , , , , .

[0050] 3. Feature localization module 300 The feature positioning module 300 includes a parameter acquisition unit 301, a feature calculation unit 302, and a wear positioning unit 303.

[0051] Parameter acquisition unit 301 is used to acquire standard data for each guide vane end face / vertical face; the standard data includes: standard propagation time for end face sealing and standard propagation time for vertical face sealing; in specific implementation, the material foundation sound velocity can be obtained from the turbine equipment technical manual, and the calibration sound velocity for each guide vane end face / vertical face sealing obtained by on-site lead breaking calibration is preferred. , and standard transmission time , .

[0052] The feature calculation unit 302 is used to calculate target feature data based on structured feature data. The target feature data includes: peak time difference matrix, amplitude ratio matrix, acoustic emission anomaly coefficient of each guide vane, vibration anomaly coefficient of each guide vane, signal arrival time difference of each guide vane, signal rise time of each guide vane, and signal dominant frequency of each guide vane. The specific feature data is as follows: (1) Determine the peak time difference matrix based on the peak time corresponding to the normalized acoustic emission peak value of each guide vane; peak time difference matrix Each element in the matrix The calculation method is as follows: , This represents the peak time corresponding to the normalized acoustic emission peak value of the i-th guide vane. The peak time corresponding to the normalized acoustic emission peak value of the j-th guide vane The difference; where , ; (2) Determine the amplitude ratio matrix based on the normalized acoustic emission peak value of each guide vane; the amplitude ratio matrix R, where each element of the matrix... The calculation method is as follows: ; Represents the normalized acoustic emission peak value of the i-th guide vane. The normalized acoustic emission peak value of the j-th guide vane The ratio of .

[0053] (3) Determine the acoustic emission anomaly coefficient based on the number of effective acoustic emission pulses for each guide vane; the acoustic emission anomaly coefficient of the i-th guide vane. ,in, This represents the number of effective acoustic emission pulses for the i-th guide vane. The arithmetic mean of the effective acoustic emission pulses of the i-th guide vane during historical fault-free events; when At that time, it was determined to be an acoustic emission anomaly. The preset acoustic emission anomaly threshold is typically set to 2.0; (4) Determine the vibration anomaly coefficient based on the normalized total vibration energy of each guide vane; the vibration anomaly coefficient of the i-th guide vane. ,in, This represents the normalized total vibrational energy of the i-th guide vane. The total vibration energy of the i-th guide vane during historical fault-free events is the arithmetic mean; when At that time, it was determined to be an abnormal vibration. The preset vibration anomaly threshold is usually set to 1.5.

[0054] (5) Signal arrival time difference : Represents the difference between the peak time of acoustic emission of the i-th guide vane and the completion time of the fully closed action; (6) Signal rise time : Represents the time it takes for the acoustic emission signal of the i-th guide vane to rise from 10% peak value to 90% peak value; (7) Signal frequency : Represents the peak frequency of the acoustic emission power spectrum of the i-th guide vane.

[0055] Wear positioning unit 303 is used to calculate positioning features based on standard data and target feature data of each guide vane end / face, employing a two-level algorithm of guide vane-level coarse positioning and sealing surface partition fine positioning, to obtain wear positioning data including faulty guide vanes and sealing types. The specific steps are as follows: (1) Coarse positioning of guide vane stage The normalized acoustic emission peak values ​​of all guide vanes are iterated to determine the maximum and second maximum normalized acoustic emission peak values. The ratio of the maximum to the second maximum normalized acoustic emission peak value is calculated to obtain the amplitude ratio. If the amplitude ratio exceeds a preset threshold, the guide vane corresponding to the maximum normalized acoustic emission peak value is directly determined to be a faulty guide vane. If the amplitude ratio does not exceed the preset threshold, cross-validation is performed by combining the maximum value of the vibration anomaly coefficient of each guide vane. In practice, all guide vanes are traversed to obtain the normalized acoustic emission peak value. Determine the guide vane number X corresponding to the maximum value; calculate the amplitude ratio. : Where S is the guide vane number of the second largest peak value; if the ratio If the ratio is directly determined to be the faulty guide vane, then X is directly identified as the faulty guide vane; if the ratio is... , combined Maximum value cross-validation: If the guide vane X, which represents the maximum normalized acoustic emission peak value, vibrates abnormally, then the guide vane is a faulty guide vane; otherwise... The largest guide vane is the faulty guide vane.

[0056] (2) Precise positioning of sealing surface zones Calculate the actual propagation time of the faulty guide vane Based on actual transmission time and end face seal standard propagation time Calculate the time difference of end face sealing Based on actual transmission time and facade sealing standard propagation time Calculate the time difference of facade sealing According to the amplitude ratio Determine the guide vane stage reliability C1; based on the end face sealing time difference. Time difference with facade sealing The difference The confidence level of the sealing surface, C2, is determined. The confidence levels of the guide vane stage, C1 and C2, are weighted and summed to obtain the positioning confidence level, Conf. When the positioning confidence level, Conf, exceeds the corresponding threshold, it is determined to be a high-confidence positioning; otherwise, it is determined to be a low-confidence positioning and marked as an anomaly to be observed.

[0057] In practice, the actual propagation time of the faulty guide vane X is calculated. ; Calculate the time difference: , ; Main frequency determination: end face seal main frequency 120–180kHz, facade sealing 80–120kHz; Ascent time determination: For the end face, For the facade.

[0058] (3) Location-based confidence quantification Conf = 0.6 × C1 + 0.4 × C2 C1: Guide vane stage reliability, ratio Take 1, Take 0.8, Take 0.5; C2: Reliability of the sealing surface Take 1, Take 0.7, Take 0.4; When Conf>0.7, it is determined to be a high-confidence location; otherwise, it is determined to be a low-confidence location and marked as an anomaly to be observed.

[0059] The data was integrated to form a wear location dataset S3, which includes faulty guide vanes and seal types. The dataset contains fields such as event ID, guide vane number, seal type, location confidence Conf, amplitude ratio, and anomaly coefficient.

[0060] 4. Diagnostic Module 400 The diagnostic module 400 includes a type analysis unit 401, a confidence assessment unit 402, and a diagnostic output unit 403.

[0061] The type analysis unit 401 is used to determine whether the type of problem is dominated by seal wear or hydraulic disturbance, by combining the timing characteristics of the acoustic emission signal, anomaly coefficients, and hydraulic parameters. The specific determination rules are as follows: (1) Sealing wear is dominant: When an acoustic emission signal appears within a specified time after the guide vane is fully closed, and the acoustic emission abnormality coefficient exceeds the corresponding threshold, it is determined to be sealing wear dominant. For example, if an acoustic emission signal appears within 0.5 seconds after the guide vane fully closes, and the acoustic emission anomaly coefficient is [not specified], then [the signal is not specified]. At this time, it is determined that the cause is seal wear. This is because the friction and impact signals generated by the wear of the sealing surface will be generated immediately at the moment the guide vane is fully closed, while the signals generated by hydraulic disturbance usually have a lag.

[0062] (2) Hydraulic disturbance is dominant: When the head fluctuation exceeds the first preset range or the tailwater level fluctuation exceeds the second preset range, and the acoustic emission anomaly coefficient is less than the corresponding threshold, it is determined to be hydraulic disturbance dominant. For example, when the head fluctuation exceeds ±5m or the tailwater level fluctuation exceeds ±1m, and the acoustic emission anomaly coefficient... At that time, it was determined that the disturbance was dominated by hydraulic disturbance.

[0063] The confidence assessment unit 402 is used to calculate the confidence level of the positioning result based on the positioning confidence level; calculate the wear intensity based on the normalized acoustic emission peak value of each guide vane and the preset moderate wear acoustic emission peak value threshold; calculate the signal anomaly degree based on the acoustic emission anomaly coefficient and the vibration anomaly coefficient; calculate the wear development trend based on the normalized acoustic emission peak value in the most recent specified events; and perform a weighted summation of the positioning result confidence level, wear intensity, signal anomaly degree, and wear development trend to obtain the comprehensive fault confidence level; the formula for calculating the comprehensive fault confidence level H is:

[0064] The meanings and calculation methods of each parameter are as follows: (1) , , , For preset weights, and Based on historical failure cases and engineering experience, the recommended value is: , , , .

[0065] (2) Indicates the reliability of the location results. .

[0066] (3) Indicates wear intensity. .in, The preset acoustic emission peak threshold for moderate wear is determined based on field calibration experiments and is typically set to 100mV. Exceed hour, Take 100.

[0067] (4) Indicates the degree of signal abnormality. .in, , For the sub-weights of acoustic emission and vibration, the recommended value is... , .

[0068] (5) This indicates the wear and tear trend, based on the last 10 events. When performing linear fitting, a slope greater than 0.05 mV / cycle indicates that wear is continuously intensifying. Take 100; otherwise, Take 0.

[0069] The diagnostic output unit 403 is used to output diagnostic conclusions and maintenance recommendations based on the overall fault confidence level and the dominant type determination results. The specific rules are as follows: (1) When the overall confidence level of the fault exceeds the first threshold and it is determined that the seal wear is the main factor, output the diagnosis result of high confidence guide vane seal wear; for example, when H≥80 and it is determined that the seal wear is the main factor, output the red warning of "high confidence guide vane seal wear" and indicate that "the end face / vertical face seal of the Xth guide vane has moderate wear, and it is recommended to arrange a shutdown for maintenance within 1 month".

[0070] (2) When the overall confidence level of the fault is between the first threshold and the second threshold, and it is determined that the seal wear is the main factor, output the diagnostic result of the medium confidence guide vane seal wear; the second threshold is less than the first threshold; for example, when 60≤H<80 and it is determined that the seal wear is the main factor, output the yellow warning of "medium confidence guide vane seal wear", indicating that "the Xth guide vane seal may have early wear, it is recommended to strengthen the inspection and pay close attention to the signal changes".

[0071] (3) When the overall confidence level of the fault is less than the second threshold (e.g., H<60) or when it is determined to be dominated by hydraulic disturbance, no warning will be issued, and only the data will be recorded for subsequent trend analysis.

[0072] Based on the above method embodiments, this application also provides a device for diagnosing wear of guide vane seals in water pumps and turbines. See [link to relevant documentation]. Figure 3As shown, the device includes: a data acquisition module 32, used to acquire multi-source data of the pump turbine guide vanes and perform time-series integration; the multi-source data includes: guide vane action signals, acoustic emission signals, vibration signals and hydraulic parameters, and the guide vane action signals include the guide vane full-closure action trigger timestamp; a feature extraction module 34, used to extract event data blocks of a specified duration from the time-series integrated multi-source data based on the guide vane full-closure action trigger timestamp, remove coupling interference caused by head, tailwater level and guide vane opening in the event data blocks through a pre-constructed hydraulic benchmark health model, and then perform feature construction to obtain structured feature data; a wear location module 36, used to perform location feature calculation on the structured feature data using a two-level algorithm of guide vane-level coarse location and sealing surface partition fine location, to obtain wear location data including faulty guide vanes and sealing types; and a diagnostic result output module 38, used to determine the dominant wear type and evaluate the comprehensive fault confidence based on hydraulic parameters, structured features and wear location data, and output graded diagnostic results.

[0073] Furthermore, the aforementioned feature extraction module 34 is used to extract data segments within a specified time period based on the trigger timestamp of the guide vane full closure action from the multi-source data after time-series integration, forming a single guide vane full closure event data block; through a pre-constructed hydraulic reference health model under normal operating conditions, interference signals in each guide vane full closure event data block are removed; key features are extracted from the guide vane full closure event data block after removing interference signals to obtain structured feature data; the structured feature data includes at least: the normalized acoustic emission peak value and corresponding peak time of each guide vane, the effective number of acoustic emission pulses of each guide vane, and the normalized total vibration energy of each guide vane.

[0074] Furthermore, the aforementioned hydraulic baseline health model includes an acoustic emission baseline health model and a vibration baseline health model, and the construction process is as follows: acquire historical fault-free operation data; divide the head into multiple intervals according to the operating range; within each interval, use a quadratic polynomial to fit the relationship between the acoustic emission signal and the guide vane opening under normal operating conditions to obtain the acoustic emission baseline health model; use a quadratic polynomial to fit the relationship between the vibration signal and the guide vane opening under normal operating conditions to obtain the vibration baseline health model; or, use a deep learning algorithm to train the model based on historical fault-free operation data to construct the acoustic emission baseline health model and the vibration baseline health model.

[0075] Furthermore, the aforementioned acoustic emission reference health model includes an acoustic emission health reference signal; the vibration reference health model includes a vibration health reference signal; the feature extraction module 34 is used to calculate the difference between the acoustic emission signal and the acoustic emission health reference signal in each guide vane fully closed event data block to obtain a normalized acoustic emission signal; calculate the difference between the vibration signal and the vibration health reference signal in the guide vane fully closed event data block to obtain a normalized vibration signal; and perform noise reduction processing on the normalized acoustic emission signal and the normalized vibration signal based on the wavelet threshold denoising method.

[0076] Furthermore, the aforementioned wear positioning module 36 is used to acquire standard data for each guide vane end face / vertical face; the standard data includes: standard propagation time of end face seal and standard propagation time of vertical face seal; based on structured feature data, target feature data is calculated; the target feature data includes: peak time difference matrix, amplitude ratio matrix, acoustic emission anomaly coefficient of each guide vane, vibration anomaly coefficient of each guide vane, signal arrival time difference of each guide vane, signal rise time of each guide vane, and signal main frequency of each guide vane; based on the standard data for each guide vane end face / vertical face and target feature data, a two-level algorithm of guide vane-level coarse positioning and sealing surface partition fine positioning is used to calculate positioning features, thereby obtaining wear positioning data including faulty guide vanes and sealing types.

[0077] Furthermore, the aforementioned wear positioning module 36 is used to determine the amplitude ratio matrix based on the normalized acoustic emission peak value of each guide vane; to determine the peak time difference matrix based on the peak time corresponding to the normalized acoustic emission peak value of each guide vane; to determine the acoustic emission anomaly coefficient based on the effective number of acoustic emission pulses of each guide vane; and to determine the vibration anomaly coefficient based on the normalized total vibration energy of each guide vane.

[0078] Furthermore, the aforementioned wear location module 36 is used to traverse the normalized acoustic emission peak values ​​of all guide vanes, determine the maximum normalized acoustic emission peak value and the second maximum normalized acoustic emission peak value; calculate the ratio of the maximum normalized acoustic emission peak value to the second maximum normalized acoustic emission peak value to obtain the amplitude ratio; if the amplitude ratio exceeds a preset threshold, the guide vane corresponding to the maximum normalized acoustic emission peak value is directly determined to be a faulty guide vane; if the amplitude ratio does not exceed the preset threshold, cross-validation is performed by combining the maximum value of the vibration anomaly coefficient of each guide vane; and the actual propagation time of the faulty guide vane is calculated. The following steps are performed: Calculate the end-face sealing time difference based on the actual propagation time and the standard propagation time for the end-face sealing; calculate the vertical sealing time difference based on the actual propagation time and the standard propagation time for the vertical sealing; determine the guide vane stage reliability based on the amplitude ratio; determine the sealing surface reliability based on the difference between the end-face sealing time difference and the vertical sealing time difference; perform a weighted summation of the guide vane stage reliability and the sealing surface reliability to obtain the positioning reliability; when the positioning reliability exceeds the corresponding threshold, it is judged as high-confidence positioning; otherwise, it is judged as low-confidence positioning and marked as abnormal and to be observed.

[0079] Furthermore, the diagnostic result output module 38 is used to determine that when an acoustic emission signal appears within a specified time after the guide vane is fully closed, and the acoustic emission anomaly coefficient exceeds the corresponding threshold, it is determined that the seal wear is dominant; when the head fluctuation exceeds the first preset range or the tailwater level fluctuation exceeds the second preset range, and the acoustic emission anomaly coefficient is less than the corresponding threshold, it is determined that the hydraulic disturbance is dominant; the positioning result reliability is calculated based on the positioning reliability; the wear intensity is calculated based on the normalized acoustic emission peak value of each guide vane and the preset moderate wear acoustic emission peak value threshold; the signal anomaly degree is calculated based on the acoustic emission anomaly coefficient and the vibration anomaly coefficient; the wear development trend is calculated based on the normalized acoustic emission peak value in the most recent specified events; the positioning result reliability, wear intensity, signal anomaly degree, and wear development trend are weighted and summed to obtain the fault comprehensive confidence; when the fault comprehensive confidence exceeds the first threshold and it is determined that the seal wear is dominant, a high-confidence guide vane seal wear diagnostic result is output; when the fault comprehensive confidence is between the first and second thresholds and it is determined that the seal wear is dominant, a medium-confidence guide vane seal wear diagnostic result is output; the second threshold is less than the first threshold.

[0080] The device provided in this application embodiment has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts of the device embodiment not mentioned can be referred to the corresponding content in the aforementioned method embodiment.

[0081] Based on the above method embodiments, this application also provides a water pump turbine guide vane seal wear diagnosis system, which includes: a server and a data acquisition system connected in communication; the data acquisition system includes multi-source sensors and a power plant computer monitoring system; the data acquisition system is used to collect multi-source data of the water pump turbine guide vane, and the server is used to execute the method described in the foregoing method embodiments.

[0082] The system provided in this application embodiment has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0083] This application also provides a computer-readable storage medium storing computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the above-described method. For specific implementation details, please refer to the foregoing method embodiments, which will not be repeated here.

[0084] The computer program products of the methods, apparatus, and electronic devices provided in the embodiments of this application include a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementations, please refer to the method embodiments, which will not be repeated here.

[0085] Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0086] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0087] In the description of this application, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0088] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The protection scope of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the technical scope disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be determined by the protection scope of the claims.

Claims

1. A method for diagnosing wear of guide vane seals in a water pump turbine, characterized in that, The method includes: The multi-source data of the guide vanes of the water pump turbine are acquired and integrated in time sequence; the multi-source data includes: guide vane action signal, acoustic emission signal, vibration signal and hydraulic parameters, and the guide vane action signal includes the guide vane full closure action trigger timestamp; Based on the trigger timestamp of the guide vane fully closing action, event data blocks of a specified duration are extracted from the multi-source data after time-series integration. The coupling interference caused by the head, tail water level and guide vane opening in the event data blocks is removed by the pre-constructed hydraulic benchmark health model, and then the structured feature data is obtained by feature construction. A two-level algorithm, namely coarse positioning at the guide vane level and fine positioning at the sealing surface partition, is used to calculate the positioning features of the structured feature data to obtain wear positioning data that includes the faulty guide vane and the sealing type. Based on the hydraulic parameters, the structured features, and the wear location data, the dominant wear type is determined and the comprehensive fault confidence is assessed, and a graded diagnosis result is output.

2. The method according to claim 1, characterized in that, Based on the trigger timestamp of the guide vane full closure action, the following steps are taken to extract event data blocks of a specified duration from the multi-source data after time-series integration: The coupling interference caused by head, tailwater level, and guide vane opening in the event data blocks is removed using a pre-constructed hydraulic benchmark health model; and then feature construction is performed to obtain structured feature data. From the multi-source data after time-series integration, data segments within a specified time period based on the trigger timestamp of the guide vane full closure action are extracted to form a single guide vane full closure event data block; By using a pre-built hydraulic baseline health model under normal operating conditions, interference signals in each guide vane full closure event data block are eliminated. Key features are extracted from the guide vane fully closed event data block after interference signals are removed to obtain structured feature data; the structured feature data includes at least: the normalized acoustic emission peak value and corresponding peak time of each guide vane, the effective number of acoustic emission pulses of each guide vane, and the normalized total vibration energy of each guide vane.

3. The method according to claim 2, characterized in that, The hydraulic baseline health model includes an acoustic emission baseline health model and a vibration baseline health model, and the construction process is as follows: Obtain historical fault-free operation data; The water head is divided into multiple zones according to its operating range; Within each interval, a quadratic polynomial is used to fit the relationship between the acoustic emission signal and the guide vane opening under normal operating conditions to obtain the acoustic emission baseline health model; a quadratic polynomial is used to fit the relationship between the vibration signal and the guide vane opening under normal operating conditions to obtain the vibration baseline health model. Alternatively, deep learning algorithms can be used to train models based on the historical fault-free operation data to construct acoustic emission benchmark health models and vibration benchmark health models.

4. The method according to claim 3, characterized in that, The acoustic emission reference health model includes an acoustic emission health reference signal; the vibration reference health model includes a vibration health reference signal. The steps for eliminating interference signals in each guide vane full closure event data block using a pre-built hydraulic baseline health model under normal operating conditions include: For each guide vane fully closed event data block, the difference between the acoustic emission signal in the guide vane fully closed event data block and the acoustic emission health reference signal is calculated to obtain the normalized acoustic emission signal; the difference between the vibration signal in the guide vane fully closed event data block and the vibration health reference signal is calculated to obtain the normalized vibration signal. The normalized acoustic emission signal and the normalized vibration signal are denoised using the wavelet threshold denoising method.

5. The method according to claim 1, characterized in that, The steps of employing a two-level algorithm—coarse positioning at the guide vane level and fine positioning in the sealing surface partition—to calculate positioning features from the structured feature data and obtain wear positioning data including the faulty guide vane and sealing type include: Acquire standard data for each guide vane end face / vertical face; the standard data includes: standard propagation time for end face sealing and standard propagation time for vertical face sealing; Based on the structured feature data, target feature data is calculated; the target feature data includes: peak time difference matrix, amplitude ratio matrix, acoustic emission anomaly coefficient of each guide vane, vibration anomaly coefficient of each guide vane, signal arrival time difference of each guide vane, signal rise time of each guide vane, and signal main frequency of each guide vane. Based on the standard data of each guide vane end face / vertical face and the target feature data, a two-level algorithm of guide vane coarse positioning and sealing surface partition fine positioning is used to calculate the positioning features, and wear positioning data including faulty guide vanes and sealing types are obtained.

6. The method according to claim 5, characterized in that, The steps for calculating target feature data based on the structured feature data include: The amplitude ratio matrix is ​​determined based on the normalized acoustic emission peak value of each guide vane; The peak time difference matrix is ​​determined based on the peak time corresponding to the normalized acoustic emission peak value of each guide vane; The acoustic emission anomaly coefficient is determined based on the number of effective acoustic emission pulses for each guide vane. The vibration anomaly coefficient is determined based on the normalized total vibration energy of each guide vane.

7. The method according to claim 5, characterized in that, Based on the standard data of each guide vane end face / vertical face and the target feature data, a two-level algorithm of guide vane-level coarse positioning and sealing surface zoning fine positioning is used to calculate positioning features, and the steps to obtain wear positioning data including faulty guide vanes and sealing types are as follows: The normalized acoustic emission peak values ​​of all guide vanes are iterated to determine the maximum and second-largest normalized acoustic emission peak values. The ratio of the maximum to the second-largest normalized acoustic emission peak values ​​is calculated to obtain the amplitude ratio. If the amplitude ratio exceeds a preset threshold, the guide vane corresponding to the maximum normalized acoustic emission peak value is directly determined to be a faulty guide vane. If the amplitude ratio does not exceed the preset threshold, cross-validation is performed by combining the maximum value of the vibration anomaly coefficient of each guide vane. Calculate the actual propagation time of the faulty guide vane; calculate the end face sealing time difference based on the actual propagation time and the standard propagation time of the end face seal; calculate the vertical sealing time difference based on the actual propagation time and the standard propagation time of the vertical seal. The confidence level of the guide vane stage is determined based on the amplitude ratio; the confidence level of the sealing surface is determined based on the difference between the end face sealing time difference and the vertical face sealing time difference; the confidence level of the guide vane stage and the confidence level of the sealing surface are weighted and summed to obtain the positioning confidence level; when the positioning confidence level exceeds the corresponding threshold, it is determined to be a high-confidence positioning; otherwise, it is determined to be a low-confidence positioning and marked as an anomaly to be observed.

8. The method according to claim 7, characterized in that, Based on the hydraulic parameters, the structured features, and the wear location data, the steps for determining the dominant wear type and assessing the comprehensive fault confidence, and outputting a graded diagnostic result, include: When an acoustic emission signal appears within a specified time after the guide vane is fully closed, and the acoustic emission abnormality coefficient exceeds the corresponding threshold, it is determined that the seal wear is the main cause. When the head fluctuation exceeds the first preset range or the tailwater level fluctuation exceeds the second preset range, and the acoustic emission anomaly coefficient is less than the corresponding threshold, it is determined that the hydraulic disturbance is dominant. Based on the positioning confidence level, the reliability of the positioning result is calculated; based on the normalized acoustic emission peak value of each guide vane and the preset moderate wear acoustic emission peak value threshold, the wear intensity is calculated; based on the acoustic emission anomaly coefficient and the vibration anomaly coefficient, the degree of signal anomaly is calculated; based on the normalized acoustic emission peak value in the most recent specified events, the wear development trend is calculated. The overall fault confidence level is obtained by weighted summing of the reliability of the positioning result, the wear intensity, the degree of signal anomaly, and the wear development trend. When the overall confidence level of the fault exceeds the first threshold and it is determined that the fault is mainly caused by seal wear, a high-confidence diagnostic result for guide vane seal wear is output. When the overall confidence level of the fault is between the first threshold and the second threshold, and it is determined that the fault is dominated by seal wear, the diagnostic result of medium confidence guide vane seal wear is output; the second threshold is less than the first threshold.

9. A device for diagnosing wear of guide vane seals in water pumps and turbines, characterized in that, The device includes: The data acquisition module is used to acquire multi-source data of the guide vanes of the water pump turbine and integrate them in time sequence; the multi-source data includes: guide vane action signal, acoustic emission signal, vibration signal and hydraulic parameters, and the guide vane action signal includes the guide vane full closure action trigger timestamp; The feature extraction module is used to extract event data blocks of a specified duration from the multi-source data after time-series integration, based on the trigger timestamp of the guide vane fully closing action. The module then removes the coupling interference caused by the head, tailwater level and guide vane opening in the event data blocks through a pre-built hydraulic benchmark health model, and then constructs features to obtain structured feature data. The wear location module is used to perform location feature calculations on the structured feature data using a two-level algorithm of coarse location at the guide vane level and fine location at the sealing surface partition, to obtain wear location data that includes the faulty guide vane and the type of seal. The diagnostic result output module is used to determine the dominant wear type and assess the comprehensive fault confidence based on the hydraulic parameters, the structured features, and the wear location data, and output the graded diagnostic results.

10. A water pump turbine guide vane seal wear diagnosis system, characterized in that, The system includes: a server and a data acquisition system connected by communication; the data acquisition system includes multi-source sensors and a power plant computer monitoring system; the data acquisition system is used to acquire multi-source data of the water pump turbine guide vanes, and the server is used to execute the method as described in any one of claims 1 to 8.

Images