A generator unified fatigue quantification monitoring method, system, device and storage medium

Abstract

The application discloses a generator unified fatigue quantification monitoring method, system, equipment and storage medium, vibration, acoustic, electrical and temperature signal data of a generator body are acquired, high-discrimination feature extraction is performed on each signal data, key physical features reflecting the state of the generator are acquired, a generator unified fatigue cumulative quantification model is established, the total fatigue value of the generator at the current moment is characterized as the combination of a basic fatigue component, an event impact component accumulation and a self-recovery component, the key physical features are mapped into standardized fatigue increments and are accumulated in the time domain, and a continuous quantification index characterizing the overall health state of the generator is obtained, an adaptive weighted fusion algorithm based on dynamic confidence is adopted, the comprehensive confidence of each data source is dynamically calculated, and the weight of each data source in health evaluation and diagnosis is adaptively allocated in combination with the correlation with the current suspected fault mode, information islands are broken, and a generator health state capable of continuous and standardized quantification is constructed.

Description

Technical Field

[0001] This invention belongs to the field of generator fault monitoring and diagnosis, and relates to a unified fatigue quantitative monitoring method, system, equipment and storage medium for generators. Background Technology

[0002] With the development of new power systems, large synchronous generators frequently participate in peak shaving and frequency regulation, and their key components face the risk of accelerated fatigue and degradation under complex operating conditions. Traditional generator condition monitoring systems mostly rely on independent threshold alarms for single physical quantities (such as vibration, temperature, and partial discharge), which has the following inherent drawbacks: Information silo problem: Each monitoring subsystem operates independently, and data and conclusions cannot be effectively correlated and complemented, making it difficult to capture early signs of failure caused by multi-physics coupling.

[0003] Lack of unified quantitative indicators: Existing monitoring mostly provides individual statuses or simple alarms, lacking a comprehensive indicator that can continuously and quantitatively characterize the overall health reserve consumption process of the generator.

[0004] Rigid fusion strategy: At the level of multi-source information utilization, there is a lack of intelligent fusion mechanism that can dynamically adjust the weight of each information source according to data quality, sensor status and specific fault modes, which affects the accuracy and robustness of diagnosis.

[0005] Insufficient early warning capability: Alarm methods based on fixed thresholds are not sensitive to the slow-accumulating aging process and weak, intermittent anomalies, making it difficult to achieve true early warning and predictive maintenance. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a unified fatigue quantitative monitoring method, system, device and storage medium for generators, break down information silos, build an intelligent monitoring framework that can continuously and standardizedly quantify the health status of generators and dynamically adjust diagnostic strategies based on data quality and fault scenarios, thereby realizing the transformation from periodic maintenance to predictive maintenance.

[0007] To achieve the above objectives, the present invention employs the following technical solution: A unified fatigue quantitative monitoring method for generators includes the following steps: Acquire vibration, acoustic, electrical, and temperature signal data of the generator body; High-discrimination feature extraction is performed on each signal data to obtain key physical features reflecting the generator state; A unified fatigue accumulation quantification model for generators is established. This model represents the total fatigue value of the generator at the current moment as a combination of basic fatigue component, event impact component accumulation and self-recovery component. Using this model, key physical characteristics are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators representing the overall health status of the generator. An adaptive weighted fusion algorithm based on dynamic confidence is adopted to dynamically calculate the comprehensive confidence of each data source and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected failure mode. Based on real-time fatigue quantification indicators and fusion diagnostic results, alarm commands are output.

[0008] Optionally, the process of extracting electrical signals with high discriminative features is as follows: extract the phase-amplitude joint statistical moment features from the partial discharge PRPD spectrum, which includes skewness and kurtosis; after converting the three-phase current to the Park vector, calculate the ripple features of the vector's mode.

[0009] Optionally, the process of high-discrimination feature extraction of vibration or audio signals is as follows: demodulate and analyze the vibration or audio signals, and extract the sideband energy ratio features centered on the line frequency and spaced at integer multiples of the rotational frequency.

[0010] Optionally, the calculation method for the basic fatigue component in the unified fatigue accumulation quantification model for generators is as follows: The calculation was performed using the Arrennis modified model based on electro-thermal dual stress, which involved aging rate coefficient, activation energy, absolute temperature of key components, operating voltage and voltage aging index. Alternatively, a natural maintenance cycle model can be used, which is based on the ratio of current operating time to the fixed maintenance time specified in the generator regulations.

[0011] Optionally, the calculation method for the event impact component and self-recovery component in the unified fatigue accumulation quantization model of the generator is as follows: The calculation of the cumulative sum of event impact components involves a normalization mapping of the abnormal event monitoring feature values. This mapping uses a nonlinear function to transform the original feature values ​​to a set range. This nonlinear function is determined by the normal baseline threshold of the feature, the feature saturation value, and the shape parameter. The calculation of the self-recovery component involves the timing of the maintenance event and the amount of fatigue recovery from a single maintenance operation. This recovery effect is set to occur instantaneously at the moment of maintenance.

[0012] Optionally, the calculation method for the overall confidence score in the adaptive weighted fusion algorithm based on dynamic confidence is as follows: The overall confidence level of the data source at the current moment is calculated. This overall confidence level is obtained by weighted summation of three factors: signal quality confidence level based on real-time signal-to-noise ratio, confidence level based on sensor self-test status, and confidence level based on the consistency of data from a group of similar sensors.

[0013] Optionally, the dynamic weight allocation method in the adaptive weighted fusion algorithm based on dynamic confidence is as follows: Once the system triggers an alert and generates a set of fault hypotheses, it determines the dynamic weight of the data source for the most likely fault hypothesis. This dynamic weight is jointly determined by the overall confidence level of the data source and the degree to which the data source's characteristics support fault diagnosis.

[0014] A unified fatigue quantification monitoring system for generators, comprising: The data acquisition module is used to acquire vibration, acoustic, electrical, and temperature signal data of the generator body. The feature extraction module is used to extract highly discriminative features from various signal data to obtain key physical features that reflect the generator's state. The characterization and mapping module is used to establish a unified fatigue accumulation quantification model for generators. This model represents the total fatigue value of the generator at the current moment as a combination of basic fatigue components, event impact components, and self-recovery components. Using this model, key physical characteristics are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators characterizing the overall health status of the generator. The adaptive weighting module is used to dynamically calculate the comprehensive confidence of each data source using an adaptive weighted fusion algorithm based on dynamic confidence, and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected fault mode. The alarm module is used to output alarm commands based on real-time fatigue quantification indicators and fusion diagnostic results.

[0015] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the generator unified fatigue quantification monitoring method.

[0016] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the generator unified fatigue quantification monitoring method.

[0017] Compared with the prior art, the present invention has the following beneficial effects: This invention fundamentally breaks down the information silos of traditional monitoring systems (vibration, electrical, temperature) by constructing a closed-loop system of perception, feature generation, quantification, fusion, and alarm. It introduces a unified fatigue accumulation quantification model, mapping heterogeneous multi-physics data to a single standard health index, solving the problem that traditional threshold alarms cannot quantify the degree of equipment health reserve depletion. Combined with a dynamic fusion strategy, it achieves a shift from point-based fault alarms to continuous full-lifecycle status assessment, significantly improving the early fault warning capability of generators and the accuracy of predictive maintenance.

[0018] Furthermore, considering the high-frequency and non-stationary characteristics of electrical signals, statistical moment and ripple analysis methods are employed to effectively improve the discriminative power of the features. Skewness and kurtosis can keenly capture the morphological distortion of partial discharge patterns, identifying insulation defect types even when the amplitude does not exceed limits; the Park vector mode ripple feature uses coordinate transformation to convert the originally concealed three-phase asymmetric faults into explicit fluctuations on the DC component. This processing method can eliminate environmental noise interference and accurately extract weak signs reflecting stator winding insulation aging and electrical asymmetry.

[0019] Furthermore, by utilizing demodulation analysis and sideband energy ratio characteristics, the challenge of extracting mechanical fault features under strong background noise was specifically addressed. Early mechanical faults in generators (such as bearing wear and air gap eccentricity) often exist in the signal in the form of modulation, which is easily masked by direct spectral analysis. By demodulating and extracting the envelope and calculating the sideband energy at specific frequency intervals, it is possible to focus on the periodic impacts caused by the fault like a microscope, thereby detecting potential mechanical problems in the rotor or bearing system in advance before the total vibration value has increased significantly.

[0020] Furthermore, the implicit aging process of the generator under normal operating conditions was quantified using the Arrhenius model or the natural cycle model. Insulating materials undergo irreversible chemical degradation under electrical and thermal stress, a fundamental type of damage that accumulates continuously over time. This calculation ensures that the fatigue model can still reflect the natural lifespan loss of the equipment even without sudden failures, providing a dynamically growing baseline for the total fatigue value and avoiding biases in health assessments caused by focusing solely on abnormal events while ignoring natural aging.

[0021] Furthermore, by introducing nonlinear mapping and a self-recovery mechanism, the nonlinear accumulation and repair process of equipment damage is realistically reproduced. The impact of abnormal shocks (such as short circuits and overloads) on lifespan exhibits a saturation effect, and the nonlinear function prevents excessive interference from single extreme values ​​on the overall assessment. Simultaneously, the introduction of a negative recovery component reflects the resetting or improvement effect of maintenance on the equipment's health status in the mathematical model. This allows monitoring indicators to dynamically follow the actual operational status of the equipment, ensuring a logical closed loop for the full lifecycle assessment.

[0022] Furthermore, by evaluating data quality through a three-dimensional approach—signal-to-noise ratio, self-test status, and group consistency—false alarms are suppressed at their source. Sensor malfunctions, electromagnetic interference, or poor contact are the main causes of false alarms in monitoring systems. Before making fusion decisions, this algorithm first performs eligibility checks on the data sources, automatically reducing the weight of low-quality or outlier sensor data. This ensures that the final diagnostic results are driven only by highly reliable data, greatly improving the system's robustness and anti-interference capabilities in complex electromagnetic environments.

[0023] Furthermore, different faults (such as inter-turn short circuits and bearing wear) exhibit vastly different sensitivities on different sensors. The adaptive weighted fusion algorithm, based on the most probable fault currently inferred, allocates decision weights to the data source most sensitive to that fault. This ensures that, when faced with a specific fault, the system automatically focuses on key information, significantly improving the accuracy of complex fault diagnosis. It achieves adaptive adjustment of the diagnostic strategy, solving the problem that fixed weights cannot accommodate multiple fault modes. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the human-computer interaction interface of the monitoring system in an embodiment of the present invention; Figure 2 This is an overview diagram of partial discharge analysis based on a flexible ultrasonic array in an embodiment of the present invention.

[0025] Figure 3 This is a detailed analysis diagram of the two sensors with the strongest partial discharge energy based on a flexible ultrasonic array in an embodiment of the present invention.

[0026] Figure 4 This is a partial discharge (PRPD) diagram based on coupling capacitor analysis in an embodiment of the present invention.

[0027] Figure 5 This is a captured image of the pulse waveform with the strongest partial discharge energy based on the coupling capacitor in an embodiment of the present invention.

[0028] Figure 6 This is a schematic diagram of the original high-frequency CT signal in an embodiment of the present invention. Detailed Implementation

[0029] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0030] The following disclosure provides many different embodiments or examples for implementing various structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed. In addition, examples of various specific processes and materials are provided in this invention, but those skilled in the art will recognize the application of other processes and / or the use of other materials.

[0031] In one embodiment of the present invention, a unified fatigue quantification monitoring method for generators is provided, comprising the following processes: Acquire vibration, acoustic, electrical, and temperature signal data of the generator body.

[0032] High-discrimination feature extraction is performed on each signal data to obtain key physical features reflecting the generator state.

[0033] The process of extracting electrical signals with high discriminative features is as follows: extract the phase-amplitude joint statistical moment features from the partial discharge PRPD spectrum, which includes skewness and kurtosis; after converting the three-phase current to the Park vector, calculate the ripple features of the vector's mode.

[0034] The process of high-discrimination feature extraction of vibration or audio signals is as follows: demodulate and analyze the vibration or audio signals, and extract the sideband energy ratio features centered on the line frequency and spaced at integer multiples of the rotational frequency.

[0035] A unified fatigue accumulation quantification model for generators is established. This model represents the total fatigue value of the generator at the current moment as a combination of basic fatigue components, event impact components, and self-recovery components. Using this model, key physical characteristics are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators that characterize the overall health status of the generator.

[0036] The calculation method for the basic fatigue component in the unified fatigue accumulation quantification model of generators is as follows: the calculation is performed using the Arrennis modified model based on electro-thermal dual stress, which involves the aging rate coefficient, activation energy, absolute temperature of key parts, operating voltage and voltage aging index; or the calculation is performed using the natural maintenance cycle model, which is based on the ratio of the current operating time to the fixed maintenance time specified in the generator regulations.

[0037] The calculation methods for the event impact component and the self-recovery component in the unified fatigue accumulation quantization model of generators are as follows: The calculation of the cumulative sum of the event impact component involves the normalization mapping of the abnormal event monitoring feature values. This mapping uses a nonlinear function to transform the original feature values ​​to a set interval. This nonlinear function is determined by the normal baseline threshold, feature saturation value and shape parameter of the feature. The calculation of the self-recovery component involves the time point of the maintenance event and the fatigue recovery amount brought about by a single maintenance. This recovery effect is set to occur instantaneously at the maintenance time.

[0038] An adaptive weighted fusion algorithm based on dynamic confidence is adopted to dynamically calculate the comprehensive confidence of each data source and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected failure mode.

[0039] The comprehensive confidence level in the adaptive weighted fusion algorithm based on dynamic confidence is calculated as follows: the comprehensive confidence level of the data source at the current moment is calculated, which is obtained by weighted summation of the signal quality confidence level based on the real-time signal-to-noise ratio, the confidence level based on the sensor self-test status, and the confidence level based on the consistency of data from a group of similar sensors.

[0040] The dynamic weight allocation method in the adaptive weighted fusion algorithm based on dynamic confidence is as follows: when the system triggers an alert and generates a set of fault hypotheses, the dynamic weight of the data source for the most likely fault hypothesis is determined. This dynamic weight is jointly determined by the comprehensive confidence of the data source and the degree to which the features of the data source support the diagnosis of faults.

[0041] Based on real-time fatigue quantification indicators and fusion diagnostic results, alarm commands are output.

[0042] In another embodiment of the present invention, a unified fatigue quantification monitoring method for generators based on multi-source information fusion is provided, comprising the following processes: S1: Construct a multi-dimensional collaborative sensing system to simultaneously collect multi-source heterogeneous monitoring data on the generator body, including vibration, acoustics, electrical properties, and temperature.

[0043] S2: Extract high-discrimination features from monitoring data from each source to obtain key physical features that reflect the equipment status.

[0044] High-discrimination feature extraction includes: Phase-amplitude joint statistical moment features, including skewness, are extracted from partial discharge PRPD maps. S qφ and kurtosis K qφ .

[0045] After converting the three-phase current to the Park vector, its magnitude is calculated. i sThe ripple characteristics are used to detect electrical asymmetry faults.

[0046] Demodulate and analyze vibration or audio signals to extract line frequencies. f line Centered on ±k× frequency f r The energy ratio characteristics of sidebands with frequency shift intervals are used to diagnose mechanical faults.

[0047] S3: Establish a unified fatigue accumulation quantification model for generators. This model will calculate the total fatigue value of the equipment at time t. F (t) represents the basic fatigue component. F b (t), cumulative event impact components and ∑ F e,i (t) and self-recovering component F r The combination of (t), i.e. Using this model, the key physical features extracted in step S2 are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators characterizing the overall health status of the equipment.

[0048] The basic fatigue component F b (t) is calculated using the Arrennis modified model based on electro-thermal dual stress, and its expression is:

[0049] In the formula, A is the aging rate coefficient. E a The activation energy is given by k, where k is the Boltzmann constant. T (τ) For the absolute temperature of critical components, V (τ) Operating voltage V 0 represents the rated voltage, and n represents the voltage aging index.

[0050] When the deployed monitoring system cannot obtain the electrical and temperature quantities of the components, basic fatigue can also be given by the natural maintenance cycle model:

[0051] In the formula, in the formula, T overhaul This refers to the fixed maintenance time specified in the equipment regulations.

[0052] The fatigue accumulation component of the event impact component F e,i (t) Calculated using the following formula:

[0053] In the formula, K s,i This is a severity coefficient used to characterize the damage weight per unit intensity for this type of event; K l,i The location coefficient is used to characterize the difference in the impact of the physical location of an event on overall health. X i (t) G( represents the monitoring characteristic value of the i-th type of abnormal event; X i (t) To make the original features X i (t) Nonlinear mapping function normalized to the [0,1] interval:

[0054] In the formula, X i,th The normal baseline threshold for this type of feature; X i,sat This is the characteristic saturation value, representing the limit of the damage growth rate after exceeding this value; m It is a shape parameter that controls the shape of the growth curve.

[0055] Self-recovering component F r Calculated using the following formula:

[0056] In the formula, t k For the first k The timing of the next maintenance event. R k This refers to the amount of fatigue recovery resulting from this maintenance (which may be partially recovered or completely reduced). δ (.) is the Dirac function, indicating that the recovery occurs instantaneously.

[0057] S4: An adaptive weighted fusion algorithm based on dynamic confidence is adopted to dynamically calculate the comprehensive confidence of each data source and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected fault mode.

[0058] The adaptive weighted fusion algorithm based on dynamic confidence specifically includes: (1) Calculate the overall confidence level of data source j at time t. C j (t) :

[0059] In the formula, Csig j (t) This is a confidence level for signal quality based on real-time signal-to-noise ratio. C sen j (t) To provide confidence levels based on the sensor's self-test status. C con j (t) To establish confidence levels based on the consistency of data from a group of similar sensors. ω 1~ ω 3 represents the weighting coefficient.

[0060] (2) When the system triggers an alert and generates a set of fault hypotheses H, for the most likely fault hypothesis h at present k Dynamic weights of data source j W j (t) for:

[0061] in, r j,k Features of data source j for diagnosing faults h k The level of support.

[0062] S5: Based on the real-time fatigue quantification index obtained in S3 and the fusion diagnosis result of step S4, execute a multi-level alarm mechanism.

[0063] The above method is implemented using the following components: The generator body and key locations are equipped with an array consisting of vibration sensors, ultrasonic sensors, noise sensors, non-contact voltage sensors, non-contact current sensors, temperature sensors, and leakage magnetic field sensors. The flexible ultrasonic sensors are arranged in a non-uniform spatial arrangement based on the electromagnetic wave propagation model of stator winding discharge to ensure that any discharge source can be captured by at least two sensors with different intensities.

[0064] The processor acquires, digitizes, and extracts high-discrimination features from the raw signals using the aforementioned sensors. The processor incorporates the unified fatigue cumulative quantization model and an adaptive weighted fusion algorithm based on dynamic confidence levels to calculate the real-time fatigue value of the device and achieve intelligent fusion diagnosis.

[0065] The processor also includes an online model parameter optimization module, which uses new observation data and maintenance feedback results within a sliding time window to perform rolling optimization of key parameters in the unified fatigue cumulative quantization model through recursive Bayesian estimation or online gradient descent.

[0066] It also features a human-machine interface connected to the processor for status visualization, alarm management, fault diagnosis report generation, and predictive maintenance decision support.

[0067] Alarm management includes: A continuous threshold range for the total fatigue value F(t) is set, corresponding to four levels: normal, attention, abnormal, and severe.

[0068] Based on the combination of the alarm status of the subsystems and the total fatigue value, multi-level comprehensive alarm rules are defined, including minor faults, moderate faults, severe faults, and emergency faults.

[0069] In another embodiment of the present invention, a complete generator health status monitoring solution is provided. Its core innovations lie in two aspects: first, it proposes a unified fatigue accumulation quantification model with clear physical meaning, mapping multi-source heterogeneous data to a unified health measurement scale; second, it designs an adaptive weighted fusion algorithm based on dynamic confidence, enabling the system to intelligently assess data reliability and focus on the most relevant fault information.

[0070] 1. Construction and implementation of a unified fatigue accumulation quantification model.

[0071] The core expression of the model is:

[0072] The expressions for each component are as follows: (1) Basic fatigue components F b (t)

[0073] In some cases, it may be impossible to obtain the correct information. T(t) and V(t) As a simplified and conservative estimate, a linear model based on natural maintenance cycles can be used:

[0074] (2) Event impact component fatigue increment F e,i (t) The calculation is as follows:

[0075] Among them, the feature normalization function G(X i (t)) :

[0076] (3) Self-recovery component F r(t), the expression is as follows:

[0077] In the formula, t k Let R be the time of the k-th maintenance event. k For the amount of fatigue recovery brought about by this maintenance (which may be partially recovered or completely zeroed), δ(.) is the Dirac function, indicating that the recovery occurs instantaneously.

[0078] 2. Multi-dimensional collaborative perception system and feature extraction implementation method.

[0079] Sensors were deployed on a 350MW water-hydrogen-cooled generator.

[0080] 3. Implementation of the adaptive weighted fusion algorithm based on dynamic confidence.

[0081] The algorithm implementation process is as follows: (1) Calculate the real-time comprehensive confidence level C j (t) :

[0082] In the formula, ω 1~ ω 3 is a fixed weight, which can be 1 / 3.

[0083] In the adaptive weighted fusion algorithm based on dynamic confidence, the comprehensive confidence of data source j at time t is... C j (t) is derived from signal quality confidence level C sig j (t) Sensor self-test confidence level C sen j (t) and group consistency confidence C con j (t) The weighted fusion is obtained; after the system enters the alert state, the final dynamic weight of data source j is... W j (t) Based on its overall confidence level C j (t) and the current most likely failure assumption h k correlation coefficient r j,k A joint decision.

[0084] Signal quality confidence C sig j(t): Evaluation based on signal-to-noise ratio (SNR). β is an adjustment parameter, which defaults to 1. Confidence drops sharply at low signal-to-noise ratios.

[0085] Sensor self-test confidence level C sen j (t): Status value (0, 1) returned based on sensor built-in diagnostics (such as zero drift, power supply monitoring).

[0086] Group consistency confidence For similar sensors, calculate the coefficient of variation (CV) of their key characteristics (such as the RMS value at power frequency). б is the standard deviation, and μ is the mean. Therefore, if a sensor's characteristic value deviates significantly from the population median, its This reduces the risk of false alarms caused by a single sensor malfunction or loose installation.

[0087] (2) Dynamic weight allocation When the system detects abnormal fatigue values ​​or issues a preliminary warning, it generates a set of suspected fault hypotheses H={ h 1 ，h 2 ，..., h m}(For example, h 1 = Rotor bar breakage, h 2 = Inter-turn short circuit). Data source j for the current dominant assumption. h k The final weight is

[0088] The weight a data source carries in decision-making depends on both its reliability ( C j (t) It also depends on its relevance to the most likely type of failure at present. r j,k ).

[0089] 4. System Deployment and Empirical Analysis.

[0090] This system was deployed on a 350MW supercritical coal-fired power unit. The application-layer human-machine interface is as follows: Figure 1 As shown, the total fatigue value, individual status, alarm information, and detailed analysis results are clearly displayed.

[0091] The system is set with four alarm threshold levels: Normal (<60%), Caution (60-80%), Abnormal (80-100%), and Severe (≥100%).

[0092] An empirical study was conducted using a partial discharge electron system as an example: Taking a Level 2 alarm event triggered by the partial radiofrequency system on a certain day as an example, this subsystem triggered a Level 2 alarm once that day using the flexible ultrasound array, while the coupling capacitance sensor and high-frequency CT monitoring were normal. The analysis data of the flexible ultrasound array for one day is as follows: Figure 2 and Figure 3 As shown (6 daily sampling trips). Figure 2 The results overview provides the ultrasonic array discharge thermogram and discharge pulse count for the day. This was obtained after functional analysis. Figure 3 The detailed analysis results shown include PRPD phase diagrams, phase amplitude distribution, and typical pulse curve captures.

[0093] The coupling capacitance sensor and high-frequency CT sensor did not trigger alarms. Functional analysis of the coupling capacitance B-phase data recorded that day yielded the following results: Figure 4 and Figure 5 The results are shown; some of the recorded high-frequency CT data are as follows: Figure 6 As shown.

[0094] The independent subsystem showed a momentary drop of 0.03% in overall health during peak load periods (natural daily average decay of 0.016%), and a slight step change in fatigue values ​​monitored by the flexible ultrasonic array. The analysis results of the data playback for this instance are as follows: (1) Confidence assessment: Flexible ultrasound signal has the highest signal-to-noise ratio, C US (t)≈0.98; the coupling capacitor signal C at this time CC When (t) drops to 0.65, the confidence level for high-frequency CT is C. CT (t) is 0.5; (2) Fault Hypothesis and Weights: The number of discharge pulses on the flexible ultrasonic steam side suddenly increased. Based on the PRPD phase distribution, a surface discharge hypothesis was given. According to matrix R, the r value of the ultrasonic feature to the event is {0.3,0.3,0.3,0.35,0.35,0.35,0.55}, of which CH07 is the manhole reference position and has the highest weight; the coupling capacitance is {0.7,0.7,0.7}; and the high-frequency CT is 0.35. (3) Adaptive weights: w in this case US (t)≈0.4、w CC (t)≈0.35、w CT (t)≈0.25. Because the surface discharge signal is weak and not easily detected by the remote system, the ultrasonic data is automatically assigned a higher weight.

[0095] Diagnostic output: After fusion, it was determined that there was intermittent surface discharge activity on the steam side of the stator winding, with a low risk level and a decline in health. The overall equipment fatigue F value did not trigger the comprehensive alarm.

[0096] Statistical analysis was conducted on all 17 valid alarm events during the two-month testing period. The system provided early warnings for all events, with an average warning lead time ΔTavg = 3.62 days. Among them, mechanical / electrical composite faults were more common. Regarding the overall false alarm rate, since no major hidden danger events were triggered and there are currently no power outage maintenance records to support this, the theoretical calculation based on the warning lead time is -23%. The overall cumulative insulation fatigue F-value is 6.917%.

[0097] Experimental results validate the effectiveness of the unified fatigue accumulation model in quantifying heterogeneous damage and achieving continuous state assessment, as well as the advantages of the dynamic confidence fusion algorithm in improving diagnostic accuracy and robustness.

[0098] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not omitted in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0099] In another embodiment of the present invention, a unified fatigue quantification monitoring system for generators is provided. This unified fatigue quantification monitoring system for generators can be used to implement the above-mentioned unified fatigue quantification monitoring method for generators. Specifically, the unified fatigue quantification monitoring system for generators includes a data acquisition module, a feature extraction module, a characterization and mapping module, an adaptive weighting module, and an alarm module.

[0100] The data acquisition module is used to acquire vibration, acoustic, electrical, and temperature signal data of the generator body.

[0101] The feature extraction module is used to extract highly discriminative features from each signal data to obtain key physical features that reflect the generator's state.

[0102] The characterization and mapping module is used to establish a unified fatigue accumulation quantification model for generators. This model represents the total fatigue value of the generator at the current moment as a combination of basic fatigue components, event impact components, and self-recovery components. Using this model, key physical characteristics are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators characterizing the overall health status of the generator.

[0103] The adaptive weighting module is used to dynamically calculate the comprehensive confidence of each data source using an adaptive weighted fusion algorithm based on dynamic confidence, and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected fault mode.

[0104] The alarm module is used to output alarm commands based on real-time fatigue quantification indicators and fusion diagnostic results.

[0105] In another embodiment of the present invention, a terminal device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can... The operation of the unified fatigue quantification monitoring method for generators includes: acquiring vibration, acoustic, electrical, and temperature signal data of the generator body; extracting high-discrimination features from each signal data to obtain key physical features reflecting the generator's state; establishing a unified fatigue accumulation quantification model for the generator, which represents the generator's total fatigue value at the current moment as a combination of basic fatigue components, event impact components, and self-recovery components; using this model to map key physical features into standardized fatigue increments and accumulate them in the time domain to obtain continuous quantitative indicators characterizing the overall health status of the generator; employing an adaptive weighted fusion algorithm based on dynamic confidence to dynamically calculate the comprehensive confidence of each data source and adaptively allocate the weight of each data source in health assessment and diagnosis based on its correlation with the current suspected fault mode; and outputting alarm commands based on real-time fatigue quantification indicators and fusion diagnostic results.

[0106] In another embodiment, the present invention also provides a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium here may include both the built-in storage medium in the terminal device and extended storage media supported by the terminal device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which may be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here may be high-speed RAM or non-volatile memory, such as at least one disk storage device.

[0107] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the unified fatigue quantification monitoring method for generators in the above embodiments. One or more instructions in the computer-readable storage medium are loaded and executed by the processor as follows: acquiring vibration, acoustic, electrical, and temperature signal data of the generator body; extracting high-discrimination features from each signal data to obtain key physical features reflecting the generator state; establishing a unified fatigue accumulation quantification model for the generator, which represents the total fatigue value of the generator at the current moment as a combination of basic fatigue component, event impact component accumulation, and self-recovery component; using this model to map key physical features into standardized fatigue increments and accumulate them in the time domain to obtain continuous quantitative indicators characterizing the overall health state of the generator; using an adaptive weighted fusion algorithm based on dynamic confidence to dynamically calculate the comprehensive confidence of each data source, and adaptively assigning the weight of each data source in health assessment and diagnosis based on its correlation with the current suspected fault mode; and outputting alarm commands based on real-time fatigue quantification indicators and fusion diagnosis results.

[0108] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0109] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0110] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0111] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0112] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0113] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0114] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0115] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0116] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

[0117] It should be understood that the above description is for illustrative purposes and not for limitation. Many embodiments and applications beyond the provided examples will be apparent to those skilled in the art upon reading the above description. Therefore, the scope of this patent should not be determined by reference to the above description, but rather by reference to the foregoing claims and the full scope of their equivalents. For purposes of completeness, all articles and references, including patent applications and publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein in the foregoing claims is not intended as a waiver of that subject matter, nor should it be construed as an indication that the applicant has not considered that subject matter as part of the disclosed inventive subject matter.

Claims

1. A unified fatigue quantitative monitoring method for generators, characterized in that, Includes the following processes: Acquire vibration, acoustic, electrical, and temperature signal data of the generator body; High-discrimination feature extraction is performed on each signal data to obtain key physical features reflecting the generator state; A unified fatigue accumulation quantification model for generators is established. This model represents the total fatigue value of the generator at the current moment as a combination of basic fatigue component, event impact component accumulation and self-recovery component. Using this model, key physical characteristics are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators representing the overall health status of the generator. An adaptive weighted fusion algorithm based on dynamic confidence is adopted to dynamically calculate the comprehensive confidence of each data source and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected failure mode. Based on real-time fatigue quantification indicators and fusion diagnostic results, alarm commands are output.

2. The generator unified fatigue quantification monitoring method according to claim 1, characterized in that, The process of extracting electrical signals with high discriminative features is as follows: extract the phase-amplitude joint statistical moment features from the partial discharge PRPD spectrum, which includes skewness and kurtosis; after converting the three-phase current to the Park vector, calculate the ripple features of the vector's mode.

3. The generator unified fatigue quantification monitoring method according to claim 1, characterized in that, The process of high-discrimination feature extraction of vibration or audio signals is as follows: demodulate and analyze the vibration or audio signals, and extract the sideband energy ratio features centered on the line frequency and spaced at integer multiples of the rotational frequency.

4. The generator unified fatigue quantification monitoring method according to claim 1, characterized in that, The calculation method for the basic fatigue component in the unified fatigue accumulation quantification model for generators is as follows: The calculation was performed using the Arrennis modified model based on electro-thermal dual stress, which involved aging rate coefficient, activation energy, absolute temperature of key components, operating voltage and voltage aging index. Alternatively, a natural maintenance cycle model can be used, which is based on the ratio of current operating time to the fixed maintenance time specified in the generator regulations.

5. The generator unified fatigue quantification monitoring method according to claim 1, characterized in that, The calculation method for the event impact component and self-recovery component in the unified fatigue cumulative quantization model of generators is as follows: The calculation of the cumulative sum of event impact components involves a normalization mapping of the abnormal event monitoring feature values. This mapping uses a nonlinear function to transform the original feature values ​​to a set range. This nonlinear function is determined by the normal baseline threshold of the feature, the feature saturation value, and the shape parameter. The calculation of the self-recovery component involves the timing of the maintenance event and the amount of fatigue recovery from a single maintenance operation. This recovery effect is set to occur instantaneously at the moment of maintenance.

6. The generator unified fatigue quantification monitoring method according to claim 1, characterized in that, The calculation method for the comprehensive confidence score in the adaptive weighted fusion algorithm based on dynamic confidence is as follows: The overall confidence level of the data source at the current moment is calculated. This overall confidence level is obtained by weighted summation of three factors: signal quality confidence level based on real-time signal-to-noise ratio, confidence level based on sensor self-test status, and confidence level based on the consistency of data from a group of similar sensors.

7. The generator unified fatigue quantification monitoring method according to claim 1, characterized in that, The dynamic weight allocation method in the adaptive weighted fusion algorithm based on dynamic confidence is as follows: Once the system triggers an alert and generates a set of fault hypotheses, it determines the dynamic weight of the data source for the most likely fault hypothesis. This dynamic weight is jointly determined by the overall confidence level of the data source and the degree to which the data source's characteristics support fault diagnosis.

8. A unified fatigue quantification monitoring system for generators, characterized in that, include: The data acquisition module is used to acquire vibration, acoustic, electrical, and temperature signal data of the generator body; The feature extraction module is used to extract highly discriminative features from various signal data to obtain key physical features that reflect the generator's state. The characterization and mapping module is used to establish a unified fatigue accumulation quantification model for generators. This model represents the total fatigue value of the generator at the current moment as a combination of basic fatigue components, event impact components, and self-recovery components. Using this model, key physical characteristics are mapped to standardized fatigue increments and accumulated in the time domain to obtain continuous quantitative indicators characterizing the overall health status of the generator. The adaptive weighting module is used to dynamically calculate the comprehensive confidence of each data source using an adaptive weighted fusion algorithm based on dynamic confidence, and adaptively allocate the weight of each data source in health assessment and diagnosis by combining its correlation with the current suspected fault mode. The alarm module is used to output alarm commands based on real-time fatigue quantification indicators and fusion diagnostic results.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the generator unified fatigue quantification monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the generator unified fatigue quantification monitoring method as described in any one of claims 1 to 7.

Images