Fault Maintenance Method for Mechanical Oil Pump Rotary Shaft of Hydro Generator Unit Based on Multi-parameter Fusion

Abstract

This application discloses a fault maintenance method for the rotating shaft of a mechanical oil pump in a hydro-generator unit based on multi-parameter fusion, comprising the following steps: acquiring in real time the operating parameters of the rotating shaft during operation, including temperature parameters, vibration parameters, and stress parameters, as well as the operating load and speed of the hydro-generator unit corresponding to the operating parameters; constructing a normal operating parameter model for the drive shaft, and determining whether the operating parameters are within the normal range based on the normal operating parameter model; constructing a fault diagnosis algorithm model, and analyzing the fault type corresponding to the operating parameters in an abnormal state based on the fault diagnosis algorithm model; constructing a multi-parameter fusion evaluation model, and analyzing the fault severity corresponding to the operating parameters in an abnormal state; and generating a corresponding maintenance plan based on the fault type and fault severity.

Description

Technical Field

[0001] This application relates to the field of hydropower equipment technology, and in particular to a fault maintenance method for the rotating shaft of a mechanical oil pump in a hydro-generator set based on multi-parameter fusion. Background Technology

[0002] As a key piece of equipment in power generation, the stable operation of hydro-generator units is crucial for ensuring the reliability of power supply. However, in the current actual operation of hydro-generator units, the mechanical oil pump drive shaft faces numerous severe challenges, with frequent failures that seriously affect the stable operation of the unit and the continuity of power production.

[0003] First, traditional fault diagnosis methods mainly rely on the monitoring and analysis of single parameters, such as focusing only on the vibration amplitude or temperature changes of the drive shaft. However, the operating environment of hydro-generator units is extremely complex, influenced by a combination of factors, and a single parameter often cannot comprehensively and accurately reflect the true operating state of the drive shaft. For example, judging solely based on vibration amplitude may overlook the potential impact of changes in other parameters such as temperature and stress on drive shaft failures. When the vibration amplitude is within the normal range, but the temperature or stress has already become abnormal, single-parameter diagnostic methods may fail to detect potential faults in time, leading to delayed fault handling, which in turn exacerbates the damage to the drive shaft and shortens its service life.

[0004] Secondly, due to the lack of comprehensive analysis integrating multiple parameters, the assessment of the severity of driveshaft failures is often inaccurate. In reality, different parameters have varying degrees of influence on the severity of a failure, and these parameters are interconnected and mutually influential. For example, in some failure scenarios, a small change in vibration amplitude may be accompanied by a sharp rise in temperature and significant fluctuations in stress; the combined changes in these parameters truly reflect the severity of the failure. However, traditional methods fail to fully consider these complex relationships, leading to either overly optimistic assessments of failure severity, resulting in a failure to take timely and effective maintenance measures, or overly conservative assessments, causing unnecessary over-maintenance and increasing maintenance costs and downtime.

[0005] Furthermore, maintenance plans based on single-parameter analysis lack comprehensiveness and specificity. Because they fail to comprehensively consider fault information reflected by multiple parameters, traditional maintenance plans may only address a single obvious fault symptom while ignoring other potential fault factors. For example, when diagnosing an imbalance fault in the drive shaft and only performing dynamic balancing, the coexisting bearing wear problem may go unnoticed. This could lead to further bearing wear during subsequent operation, triggering new faults and affecting the normal operation of the unit.

[0006] Furthermore, as hydro-generator units develop towards larger capacity and higher parameters, the loads and working environments borne by mechanical oil pump drive shafts are becoming increasingly complex. Traditional fault diagnosis and maintenance methods are finding it increasingly difficult to meet the reliability and stability requirements of modern hydro-generator units. For example, in some new large-capacity hydro-generator units, the drive shaft rotates at higher speeds and bears greater pressures, leading to a corresponding increase in the probability and complexity of faults. In such cases, traditional methods relying on single parameters prove inadequate when facing complex and ever-changing fault modes.

[0007] In summary, existing methods for diagnosing and maintaining mechanical oil pump drive shafts have many shortcomings and cannot meet the ever-increasing demand for stable operation of hydro-generator units. Therefore, there is an urgent need for a fault maintenance method based on multi-parameter fusion to comprehensively and accurately diagnose faults, precisely assess their severity, and develop scientific, reasonable, and targeted maintenance plans. This would improve the reliability and service life of mechanical oil pump drive shafts, ensure the stable operation of hydro-generator units, and reduce maintenance costs and power generation risks. Summary of the Invention

[0008] The main objective of this application is to provide a fault maintenance method for the mechanical oil pump drive shaft of a hydro-generator unit based on multi-parameter fusion, so as to solve the problem that there are many shortcomings in the existing mechanical oil pump drive shaft fault diagnosis and maintenance methods, which cannot meet the growing demand for stable operation of hydro-generator units.

[0009] To achieve the above objectives, this application provides the following technical solution: A method for fault maintenance of the rotating shaft of a mechanical oil pump in a hydro-generator unit based on multi-parameter fusion includes the following steps: The operating parameters of the rotating shaft during operation are acquired in real time, including temperature parameters, vibration parameters, and stress parameters, as well as the operating load and speed of the hydro-generator unit corresponding to the operating parameters. Construct a normal operating parameter model for the drive shaft, and determine whether the operating parameters are within the normal range based on the normal operating parameter model; Construct a fault diagnosis algorithm model, and analyze the fault type corresponding to the operating parameters in an abnormal state based on the fault diagnosis algorithm model; Construct a multi-parameter fusion evaluation model to analyze the severity of faults corresponding to operating parameters in abnormal states; A corresponding maintenance plan is generated based on the fault type and severity.

[0010] As a further improvement to this application, the operating parameters of the rotating shaft during operation are acquired in real time, including the following steps: Triaxial vibration sensors are installed at the bearing seats near both ends of the drive shaft to accurately measure vibration parameters along the X, Y, and Z directions, including vibration amplitude, vibration frequency, and vibration phase. A surface temperature sensor is used on the drive shaft to collect the surface temperature of the drive shaft in real time. Strain gauge stress sensors are installed at the connection points of the coupling. By detecting the minute strain generated when the coupling is subjected to force, the stress is converted into an electrical signal to measure the magnitude of the stress.

[0011] As a further improvement to this application, a normal operating parameter model of the drive shaft is constructed, including the following steps: Normal operating parameters of the drive shaft under different working conditions are collected under normal operating conditions, including normal temperature parameters, normal vibration parameters and normal stress parameters under different load and speed conditions. The collected normal operation parameters are processed for time synchronization. The collected normal operation parameters are subjected to feature extraction and transformation to mine representative and discriminative feature variables; The feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions; The parameter models under different operating conditions are integrated to form a complete normal operation parameter model library.

[0012] As a further improvement to this application, feature extraction and transformation are performed on the collected normal operation parameters to mine representative and discriminative feature variables, including the following steps: To mine the characteristic variables of normal temperature parameters, the following steps are performed: Extract basic temperature features, including the current temperature value, the rate of temperature change, and the temperature mean and variance; Extract temperature trend features, including long-term temperature trend features and seasonal temperature change features; For the mining of characteristic variables of normal vibration parameters, the following steps are performed: Extract the time-domain features of normal vibration parameters, including vibration amplitude, peak factor, and kurtosis; Extract the frequency domain characteristics of normal vibration parameters, including spectral analysis features, power spectral density, and frequency centroid; For characteristic variable mining of normal stress parameters, the following steps are performed: Extract stress magnitude characteristics, including stress amplitude, average stress, and stress ratio; Extract stress cycle characteristics, including the number of stress cycles and cyclic stress characteristic parameters.

[0013] As a further improvement to this application, the feature extraction and transformation of the collected normal operation parameters to mine representative and discriminative feature variables also includes the following steps: Extract load-related features, including the correlation between load and normal vibration parameters, and the relationship between load and normal temperature parameters; Extract rotational speed-related features, including the matching between rotational speed and normal vibration frequency, and the relationship between rotational speed and normal temperature and stress parameters.

[0014] As a further improvement to this application, the feature variables are input into an optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions, including the following steps: The feature variables are organized into a dataset in a unified format. Each data record in the dataset contains information on multiple dimensions of the drive shaft at a certain moment, including vibration amplitude, frequency characteristics, temperature value, temperature change rate, stress magnitude, and stress ratio. Based on the characteristics of data distribution and the needs of practical applications, a suitable clustering algorithm is determined to perform clustering analysis on the data in the dataset. The quality of clustering is evaluated by calculating the silhouette coefficient and CH index of the clustering results. Perform statistical analysis on each clustering result and calculate the statistical parameters of the mean, median, standard deviation, minimum, and maximum of each feature variable; Based on the statistical parameters of each clustering result, the normal operating parameter range of the drive shaft under different working conditions is determined.

[0015] As a further improvement to this application, the feature variables are input into an optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions, and the following steps are also included: By calculating the correlation coefficients between various feature variables, the linear or nonlinear correlation between different feature variables can be determined. Based on the results of feature correlation analysis, a feature relationship model among the feature variables is constructed; By integrating the normal operating parameter ranges and characteristic relationship models under different working conditions, a complete normal operating parameter model for the drive shaft is formed.

[0016] As a further improvement to this application, a fault diagnosis algorithm model is constructed, and the fault type corresponding to the operating parameters in an abnormal state is analyzed based on the fault diagnosis algorithm model, including the following steps: A fault diagnosis algorithm model is established based on a neural network architecture, and the operating parameters in an abnormal state are input into the fault diagnosis algorithm model. The current operating status of the drive shaft is determined by performing feature extraction and pattern matching on the input data. When a fault is detected in the drive shaft, the location of the fault is determined by analyzing the propagation path and weight distribution of each data acquisition point in the neural network.

[0017] As a further improvement to this application, a multi-parameter fusion evaluation model is constructed to analyze the fault severity corresponding to the operating parameters in an abnormal state, including the following steps: For each type of operating parameter related to drive shaft failure, multiple specific parameter indicators are selected to assess the severity of the failure. The weight of each parameter index in explaining the severity of the fault was determined by weight analysis. Multiply the actual value of each parameter by its corresponding weight, and then sum the weighted values ​​of all parameters to obtain a comprehensive fault severity index.

[0018] As a further improvement to this application, a corresponding maintenance plan is generated based on the fault type and severity, including the following steps: For minor driveshaft imbalance faults, dynamic balancing tests and corrections can be performed on the driveshaft during the next scheduled maintenance, and detailed test procedures and standards will be provided. For more complex or rare faults, deep learning algorithms are combined to analyze a large number of historical fault cases and solutions to generate targeted maintenance plans. During the scheme generation process, the optimal maintenance strategy is found by taking into account maintenance costs, maintenance time, and factors affecting unit operation.

[0019] The beneficial effects of this application are as follows: (1) By integrating multiple parameters such as vibration, temperature, and stress for comprehensive analysis, the limitations of traditional single-parameter diagnosis are overcome, enabling a more comprehensive judgment of the fault type and cause. This avoids overlooking other potential fault factors due to a single normal parameter, greatly improving the accuracy of fault diagnosis. It effectively avoids unnecessary maintenance or delayed fault handling caused by misjudgment.

[0020] (2) This method can accurately assess the severity of faults by considering the different impacts of various parameters on the severity of the fault and the interrelationships between these parameters. The weights of each parameter are determined through a weighted analysis algorithm, resulting in a more realistic fault severity index. This allows maintenance personnel to allocate maintenance resources and strategies rationally based on accurate severity assessments, avoiding over-maintenance or under-maintenance. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the steps of the fault maintenance method for the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion in this application. Figure 2 This is a flowchart illustrating the method steps for acquiring the operating parameters of the rotating shaft in real time during step S1 of this application. Figure 3 This is a flowchart illustrating the method steps for constructing the normal operation parameter model of the drive shaft in step S2 of this application. Figure 4 This is a flowchart illustrating the method for mining characteristic variables of normal temperature parameters in step S23 of this application. Figure 5 This is a flowchart illustrating the method steps for mining characteristic variables of normal vibration parameters in step S23 of this application. Figure 6 This is a flowchart illustrating the method steps for mining characteristic variables of normal stress parameters in step S23 of this application. Figure 7 This is a flowchart illustrating the preferred method for extracting and transforming features from the collected normal operation parameters and mining representative and discriminative feature variables in step S23 of this application. Figure 8 This is a flowchart illustrating the method steps in step S24 of this application, where the feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions. Figure 9 The flowchart of the method for inputting the feature variables into the optimized clustering algorithm for cluster analysis in step S24 of this application to obtain the optimal parameter model under different working conditions is shown below. Figure 10 This is a flowchart illustrating the method steps for constructing a fault diagnosis algorithm model and analyzing the fault type corresponding to the operating parameters in an abnormal state based on the fault diagnosis algorithm model in step S3 of this application. Figure 11 This is a flowchart illustrating the method steps for constructing a multi-parameter fusion evaluation model and analyzing the severity of faults corresponding to operating parameters in an abnormal state in step S4 of this application. Figure 12 This is a flowchart illustrating the method steps in step S5 of this application for generating a corresponding maintenance plan based on the fault type and fault severity. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0023] This application discloses a multi-parameter fusion-based method for maintaining the faults of the mechanical oil pump rotating shaft in a hydro-generator unit. This method is primarily applied to the maintenance of the mechanical oil pump rotating shaft in hydro-generator units, extending the service life of the drive shaft. The mechanical oil pump, as a core component of the lubrication and hydraulic system of a hydro-generator unit, is responsible for providing necessary lubricating oil and pressurized oil to various parts of the unit, ensuring its normal operation. The mechanical oil pump drive shaft, as a key component driving the pump, directly affects the performance and reliability of the mechanical oil pump and even the entire hydro-generator unit.

[0024] like Figure 1 As shown, the fault maintenance method for the rotating shaft of the mechanical oil pump of a hydro-generator unit based on multi-parameter fusion provided in this application includes the following steps: S1. Real-time acquisition of the operating parameters of the rotating shaft during operation, including temperature parameters, vibration parameters, and stress parameters, as well as the operating load and speed of the hydro-generator unit corresponding to the operating parameters.

[0025] S2. Construct a normal operating parameter model for the drive shaft, and determine whether the operating parameters are within the normal range based on the normal operating parameter model.

[0026] S3. Construct a fault diagnosis algorithm model, and analyze the fault type corresponding to the operating parameters in an abnormal state based on the fault diagnosis algorithm model.

[0027] S4. Construct a multi-parameter fusion evaluation model to analyze the severity of the fault corresponding to the operating parameters in an abnormal state.

[0028] S5. Generate a corresponding maintenance plan based on the fault type and severity.

[0029] In the above steps, by integrating multiple parameters such as vibration, temperature, and stress, fault information is comprehensively captured. Then, by constructing a normal operating parameter model and a fault diagnosis algorithm model, combined with the comprehensively collected operating parameters of the drive shaft, diagnostic accuracy is significantly improved, and the misdiagnosis rate is significantly reduced. By comprehensively considering the impact of different parameters on the severity of the fault and the correlation between parameters, scientific methods are used to determine weights, achieving an accurate assessment of the fault severity and providing a reliable basis for rationally planning maintenance strategies. Based on the results of multi-parameter fusion diagnosis and assessment, a scientific, reasonable, and targeted maintenance plan is formulated, not only addressing fault symptoms but also deeply analyzing the root causes, effectively preventing fault recurrence, and extending the service life of the drive shaft.

[0030] like Figure 2As shown, in step S1, the operating parameters of the rotating shaft during operation are acquired in real time, including the following steps: S11. Install triaxial vibration sensors at the bearing housings near both ends of the drive shaft to accurately measure vibration parameters along the X, Y, and Z directions, including vibration amplitude, vibration frequency, and vibration phase. This type of sensor was chosen because of its high sensitivity and wide-bandwidth response characteristics, which can capture extremely subtle vibration changes in the drive shaft during operation. These changes are often early signs of potential faults. For example, when the drive shaft becomes unbalanced due to installation misalignment or component wear, the vibration sensor can promptly detect an abnormal increase in vibration amplitude within a specific frequency range.

[0031] S12. A contact-type temperature sensor is used on the surface of the drive shaft to collect the surface temperature of the drive shaft in real time. An abnormal rise in temperature may indicate problems such as poor lubrication, increased friction of components, or cooling system failure. By monitoring the temperature data in real time, these potential problems can be detected in a timely manner. The contact-type temperature sensor has good thermal conductivity and fast response capability, and can accurately measure the surface temperature of the drive shaft in real time. Considering the complexity of the operating environment of the hydro-generator set, a temperature sensor with high temperature resistance, moisture resistance, and electromagnetic interference resistance can be selected to ensure stable operation even in harsh environments.

[0032] S13. Strain gauge stress sensors are installed at the connection points of the coupling. These sensors detect minute strains generated when the coupling is under stress and convert them into electrical signals to measure the stress magnitude. As a crucial component for power transmission, the coupling endures complex stresses during operation. Stress sensors can monitor stress changes in the coupling in real time. When the stress exceeds the normal range, it indicates potential problems such as loosening, deformation, or fatigue damage, providing a basis for timely maintenance.

[0033] like Figure 3 As shown, in step S2, constructing the normal operating parameter model of the drive shaft includes the following steps: S21. Under normal operating conditions, collect normal operating parameters of the drive shaft under different working conditions, including normal temperature parameters, normal vibration parameters, and normal stress parameters under different load and speed conditions.

[0034] S22. Perform time synchronization processing on the collected normal operation parameters.

[0035] S23. Extract and transform features from the collected normal operation parameters to discover representative and discriminative feature variables.

[0036] S24. Input the feature variables into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions.

[0037] S25. Integrate the parameter models under different working conditions to form a complete normal operation parameter model library.

[0038] In step S21, data closely related to the operating status of the drive shaft can be extracted from the historical normal operation data records of the hydro-generator unit. This data covers monitoring records from various sensors under different operating conditions over a long period. This data includes not only vibration amplitude, frequency, and phase information collected by vibration sensors, drive shaft surface temperature and temperature change rate recorded by temperature sensors, and direct parameters such as stress magnitude and stress change trends obtained by stress sensors, but also operating condition data such as the operating load and speed of the hydro-generator unit, ensuring the completeness and comprehensiveness of data collection. Because the operating status of the drive shaft is closely related to the overall operating condition of the unit, the normal parameter range of the drive shaft will differ under different load and speed conditions.

[0039] In step S22, the collected normal operation parameters are time-synchronized. This ensures that data from different sensors at the same point in time accurately correspond and reflect the overall operating status of the drive shaft at the same moment. Since the sampling frequencies of each sensor may differ, existing interpolation or resampling methods can be used to unify all data to the same time scale for subsequent analysis.

[0040] In step S23, feature extraction and transformation are performed on the collected normal operation parameters to mine representative and discriminative feature variables, including the following steps: like Figure 4 As shown, the following steps are performed to mine the characteristic variables of normal temperature parameters: S2311. Extract basic temperature features, including the current temperature value, the rate of temperature change, and the temperature mean and variance.

[0041] For the current temperature value, the real-time temperature of the drive shaft surface measured by the temperature sensor is directly recorded. This is the most intuitive temperature feature and can reflect the current heat generation of the drive shaft.

[0042] The rate of temperature change is calculated by dividing the temperature difference between adjacent time points by the time interval. This rate reflects the speed at which temperature rises or falls, and is crucial for assessing the stability of equipment operation. For example, a sudden, rapid temperature rise may indicate poor lubrication or increased friction in components, while a slow temperature rise may be a normal process of heat accumulation.

[0043] For temperature mean and variance, the average and variance of the temperature can be calculated over a period of time. The mean reflects the average temperature level of the drive shaft during that period, while the variance reflects the degree of temperature fluctuation. Under stable operating conditions, the temperature mean should remain within a certain range, and the variance should be small; if the variance increases, it indicates that the temperature fluctuation is large, which may be due to factors such as unstable equipment operation or external environmental interference.

[0044] S2312. Extract temperature trend features, including long-term temperature trend features and seasonal temperature change features.

[0045] For long-term temperature trends, existing techniques such as multinomial fitting or time series analysis can be used to analyze temperature data over a longer period. By fitting the curve, the temperature change trend over a future period can be predicted, allowing for the early detection of potential temperature anomalies. For example, if the fitted curve shows a continuous upward trend in temperature, even if the current temperature is still within the normal range, it should be taken seriously, and the cause should be further investigated.

[0046] Regarding seasonal temperature variation characteristics, considering that the operating environment and workload of the hydro-generator unit may change with the seasons, the analysis focuses on the variation patterns of temperature data across different seasons. For example, in summer when the ambient temperature is higher, the temperature of the drive shaft may generally be higher than in winter, but within each season, the temperature variation should exhibit a relatively stable pattern. By extracting seasonal temperature variation characteristics, the normal operating temperature range of the drive shaft under different seasonal conditions can be determined more accurately.

[0047] like Figure 5 As shown, the following steps are performed to mine the characteristic variables of normal vibration parameters: S2321. Extract the time-domain characteristics of normal vibration parameters, including vibration amplitude, peak factor and kurtosis.

[0048] For vibration amplitude, the vibration amplitude measured by the vibration sensor can be directly obtained, which is a fundamental characteristic describing vibration intensity. Not only is the instantaneous value of the vibration amplitude recorded, but the maximum, minimum, mean, and root mean square (RMS) values ​​over a period of time are also calculated. The maximum value reflects the peak intensity during the vibration process, the minimum value can be used to determine if there is abnormally low vibration, the mean value reflects the average level of vibration, and the RMS value more accurately reflects the magnitude of vibration energy, which is of great significance for assessing the overall vibration condition of the drive shaft.

[0049] The crease factor can be obtained by calculating the ratio of the maximum vibration amplitude to the root mean square value. The crease factor highlights the impact component in the vibration signal. When a local fault occurs in the drive shaft, such as a crack or wear, it will cause impact pulses in the vibration signal, increasing the crease factor. Therefore, the crease factor is one of the sensitive features for detecting potential faults.

[0050] Kurtosis measures the steepness of the amplitude distribution of a vibration signal. During normal operation, the vibration signal amplitude distribution is relatively stable, and the kurtosis value is relatively stable. However, when a drive shaft malfunctions, abnormal impacts or periodic components appear in the signal, leading to an increase in the kurtosis of the amplitude distribution. By monitoring changes in kurtosis, abnormalities in the drive shaft's operating condition can be effectively identified.

[0051] S2322. Extract the frequency domain characteristics of normal vibration parameters, including spectral analysis characteristics, power spectral density, and frequency centroid.

[0052] For spectral analysis, the time-domain vibration signal can be converted into a frequency-domain signal using the Fast Fourier Transform (FFT) technique, which provides the distribution of vibration amplitude with frequency. Analyzing the main frequency components and their corresponding amplitudes in the spectrum helps determine the characteristic frequencies of the drive shaft during normal operation. For example, since the rotational frequency and harmonic frequencies of the drive shaft are relatively stable during normal operation, monitoring the amplitude changes of these frequency components can determine whether the drive shaft's rotational speed is stable and whether there are frequency shifts or new frequency components caused by imbalance, misalignment, or other problems.

[0053] Power spectral density (PSD) represents the signal power per unit frequency band, providing a clearer picture of the distribution of vibration energy across different frequencies. Calculating PSD helps determine which frequency components contribute significantly to vibration energy and how energy distribution changes under different operating conditions. For example, under certain fault conditions, the PSD may increase significantly within a specific frequency range; analyzing the PSD allows for more accurate localization of fault-related frequency characteristics.

[0054] The center of gravity of a frequency refers to the average frequency of all frequency components in the spectrum, weighted according to their corresponding amplitudes. It reflects the position of the center of gravity of vibration energy on the frequency axis. When the operating state of the drive shaft changes, the center of gravity of the frequency will also change accordingly. For example, in the process of the gradual development of an imbalance fault, the center of gravity of the frequency may move towards a specific frequency region. By monitoring the changes in the center of gravity of the frequency, potential fault trends can be detected in advance.

[0055] like Figure 6 As shown, the following steps are performed for feature variable mining of normal stress parameters: S2331. Extract stress magnitude characteristics, including stress amplitude, average stress, and stress ratio.

[0056] The stress amplitude refers to the maximum and minimum stress values ​​measured by the stress sensor. The stress amplitude reflects the range of stress variations experienced by the drive shaft during operation. Excessive stress amplitude can lead to fatigue damage to the drive shaft, shortening its service life.

[0057] Mean stress is calculated by averaging the stress over a period of time, reflecting the average load level borne by the drive shaft. Under prolonged conditions of high mean stress, the drive shaft is more prone to plastic deformation and fatigue failure.

[0058] The stress ratio, defined as the ratio of minimum stress to maximum stress, has a significant impact on the fatigue life of materials. Different stress ratios correspond to different fatigue damage mechanisms, and by monitoring changes in the stress ratio, the degree of fatigue damage to the drive shaft can be assessed.

[0059] S2332. Extract stress cycle characteristics, including the number of stress cycles and cyclic stress characteristic parameters.

[0060] The stress cycle count is a statistical measure of the number of stress cycles within a given time period. It is a crucial parameter for assessing the fatigue life of a drive shaft. Frequent stress cycles accelerate material fatigue damage. By monitoring the stress cycle count in real time, the remaining fatigue life of the drive shaft can be predicted.

[0061] Characteristic parameters of cyclic stress, such as the standard deviation of stress amplitude and the asymmetry coefficient of stress cycle, can more comprehensively describe the characteristics of stress cycle and provide richer information for fatigue analysis.

[0062] like Figure 7 As shown, in step S23, feature extraction and transformation are performed on the collected normal operation parameters to mine representative and discriminative feature variables. This also includes the following steps: S234. Extract load-related features, including the correlation between load and normal vibration parameters, and the relationship between load and normal temperature parameters.

[0063] The correlation between load and normal vibration parameters can be analyzed by examining changes in vibration amplitude, frequency, and other characteristics as the load increases. By establishing a load-vibration relationship curve, it is possible to more accurately determine whether the drive shaft vibration is normal under different load conditions. If the vibration characteristics deviate from the normal load-vibration relationship curve when the load changes, it may indicate a potential fault.

[0064] Regarding the relationship between load and normal temperature parameters, as the load increases, the load on the drive shaft increases, which may lead to a rise in temperature. By analyzing the slope and trend of the load-temperature curve, the heat dissipation performance and operational stability of the drive shaft under different loads can be evaluated. If the temperature rises too quickly or exceeds the normal load-temperature range within a certain load range, it may indicate problems such as poor heat dissipation or excessive component friction.

[0065] S235. Extract rotational speed-related features, including the matching between rotational speed and normal vibration frequency, and the relationship between rotational speed and normal temperature and stress parameters.

[0066] Regarding the matching between rotational speed and normal vibration frequency, the vibration frequency of the drive shaft is closely related to its rotational speed. Under normal circumstances, the vibration frequency should correspond to a specific multiple of the rotational speed, such as the first harmonic, second harmonic, etc. By monitoring the matching between vibration frequency and rotational speed, it is possible to determine whether there are faults such as imbalance or misalignment in the drive shaft. For example, when abnormal non-integer harmonic components are found in the vibration frequency, it may be due to structural defects or installation problems in the drive shaft.

[0067] Regarding the relationship between rotational speed and normal temperature and stress parameters, when the rotational speed changes suddenly, the drive shaft will be subjected to additional inertial forces and torque, which may lead to a sudden increase in temperature and a sharp change in stress. Analyzing the response characteristics of these parameters during rotational speed changes can provide a more comprehensive understanding of the drive shaft's operating status under dynamic conditions and help identify potential failure risks in a timely manner.

[0068] Through the comprehensive and meticulous feature engineering process described above, rich and representative feature variables are extracted from the raw sensor data. This provides strong support for constructing an accurate and reliable normal operation parameter model for the driveshaft, thus laying a solid foundation for the subsequent construction of a fault diagnosis algorithm model. In practical applications, these features will help the model more accurately identify the normal operating state and potential fault modes of the driveshaft, improving the accuracy and reliability of fault diagnosis.

[0069] like Figure 8 As shown, in step S24, the feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions, including the following steps: S241. Organize the feature variables into a dataset according to a unified format. Each data record in the dataset contains information on multiple dimensions of the drive shaft at a certain moment, including vibration amplitude, frequency characteristics, temperature value, temperature change rate, stress magnitude, and stress ratio.

[0070] S242. Based on the data distribution characteristics and actual application requirements, determine a suitable clustering algorithm to perform clustering analysis on the data in the dataset.

[0071] S243. The quality of clustering is evaluated by calculating the silhouette coefficient and CH index of the clustering results.

[0072] S244. Perform statistical analysis on each clustering result and calculate the statistical parameters of the mean, median, standard deviation, minimum, and maximum values ​​of each feature variable.

[0073] S245. Based on the statistical parameters of each clustering result, determine the normal operating parameter range of the drive shaft under different working conditions.

[0074] In step S241, routine data cleaning is performed on the feature variable data to remove noise, outliers, and duplicate data. Then, feature variables with different dimensions and value ranges are normalized, using methods such as max-min normalization or Z-fraction normalization. This scales the data to a uniform, fixed interval, making data of different magnitudes comparable. Next, time synchronization processing is performed on the feature variable data to ensure accurate correspondence between data from different sensors at the same time point, reflecting the overall operating status of the drive shaft at the same moment.

[0075] After the aforementioned feature extraction, data cleaning, normalization, and time synchronization processing, all feature variable data are organized into a dataset according to a unified format. This dataset contains various feature variables reflecting different operating characteristics of the drive shaft, obtained from multiple sensors such as vibration, temperature, and stress sensors. Each data record contains information on multiple dimensions of the drive shaft at a specific moment, including vibration amplitude, frequency characteristics, temperature value, temperature change rate, stress magnitude, and stress ratio. For example, a data record at a specific moment t1 might be represented as: [Vibration amplitude] t1 Frequency amplitude t1 Frequency and phase t1 Temperature value t1 Temperature change rate t1 Stress magnitude t1 stress ratio t1 ,...].

[0076] This dataset provides a comprehensive, accurate, and consistent data foundation for subsequent cluster analysis, model training, and fault diagnosis.

[0077] In step S242, based on the data distribution characteristics and practical application requirements, the DBSCAN clustering algorithm, commonly used in data clustering analysis, is selected to perform clustering analysis on the data in the dataset. Specifically, for each data point in the dataset, a circle is drawn with that point as the center and the neighborhood radius eps as the radius. The number of data points within the circle, including the point itself, is counted. If the number of data points within the circle is greater than or equal to the minimum number of points minPts, then that point is determined to be a core point. For example, if a sufficient number of data points within the neighborhood of a data point in the dataset's eps meet the minPts requirement, that point is a core point. Starting from a core point, all points within its eps neighborhood, including both core and non-core points, are grouped into the same cluster. For core points among these neighborhood points, its neighborhood is further expanded, including newly discovered neighborhood points, continuously expanding the cluster range until it can no longer be expanded. In this way, a series of density-connected data point sets are formed, completing the data clustering analysis.

[0078] In step S243, the silhouette coefficient is used to evaluate the quality of clustering. It comprehensively considers both the tightness of a data point with its own cluster and its separation from other clusters. For each data point x... i First, calculate the average distance a(x) between it and other data points within the same cluster. i This reflects how closely a data point is clustered within its own cluster; then, the minimum average distance b(x) between it and data points in other clusters is calculated. i This reflects the degree of separation between this cluster and other clusters. Silhouette coefficient s(x) i The formula for calculating ) is: The silhouette coefficient ranges from -1 to 1. The closer it is to 1, the better the clustering effect, that is, the data points are closely packed within their respective clusters and have a high degree of separation from other clusters.

[0079] The CH index evaluates clustering effectiveness by calculating the ratio of between-cluster dispersion to within-cluster dispersion. Higher between-cluster dispersion and lower within-cluster dispersion result in a higher CH index and better clustering performance. It helps determine whether the clustering results clearly delineate different operating condition categories.

[0080] In step S244, for each clustering result, the statistical parameters such as mean, median, standard deviation, minimum, and maximum of each feature variable are calculated. For example, for a certain cluster, assuming it represents a high-load operating condition, the mean of the vibration amplitude is calculated to be μ. vibration The standard deviation is σ vibration The average temperature is μ temperature The median stress is M. stress These statistical parameters reflect the central tendency and dispersion of the various operating parameters of the drive shaft under this specific working condition.

[0081] In step S245, based on the statistical parameters of each clustering result, the normal operating parameter range of the drive shaft under different working conditions is determined. Generally, the mean ± k times the standard deviation can be used, where k is determined according to the actual situation, and is usually taken as 1-3 as the fluctuation range of the normal operating parameters. For example, the normal operating range of vibration amplitude can be set to [μ vibration 2σ vibration μ vibration +2σ vibration For some parameters that have a significant impact on the safe operation of the equipment, the range may be further adjusted based on the minimum and maximum values ​​to ensure that the parameters do not exceed the safety limits during normal operation.

[0082] Through the detailed cluster analysis and model building process described above, an accurate and comprehensive normal operating parameter model of the hydro-generator drive shaft can be established, providing a solid foundation for subsequent fault diagnosis, performance evaluation, and other work based on this model.

[0083] like Figure 9 As shown, in step S24, the feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions. This also includes the following steps: S246. By calculating the correlation coefficients between various characteristic variables, the linear or nonlinear correlation between different characteristic variables can be determined. Specifically, the degree of linear or nonlinear correlation between different characteristics can be determined by calculating methods such as Pearson correlation coefficient and Spearman correlation coefficient. For example, it may be found that temperature and vibration amplitude have a positive correlation under certain working conditions, that is, as the temperature increases, the vibration amplitude also tends to increase; or there may be a certain functional relationship between stress and rotational speed. A correlation matrix or scatter plot can be drawn to visually display the relationship between the various characteristics.

[0084] S247. Based on the results of feature correlation analysis, construct a feature relationship model among the feature variables. For linearly correlated features, a linear regression model can be used to describe their relationship; for non-linearly correlated features, methods such as multinomial regression and neural network regression may be used. For example, establish a linear regression equation V between temperature and vibration amplitude. ibration =a×T+b, where a and b are coefficients determined through regression analysis. These characteristic relationship models further enrich the content of normal operation parameter models, enabling them not only to describe the normal range of individual parameters but also to reflect the intrinsic relationships between parameters.

[0085] S248. Integrate the normal operating parameter ranges and characteristic relationship models under different working conditions to form a complete normal operating parameter model for the drive shaft. This model comprehensively describes the normal value ranges of each parameter of the drive shaft under different operating conditions and the interrelationships between the parameters, providing an accurate reference for real-time monitoring of the drive shaft's operating status.

[0086] like Figure 10 As shown, in step S3, a fault diagnosis algorithm model is constructed, and the fault type corresponding to the operating parameters in an abnormal state is analyzed based on the fault diagnosis algorithm model, including the following steps: S31. Establish a fault diagnosis algorithm model based on a neural network architecture, and input the operating parameters in an abnormal state into the fault diagnosis algorithm model.

[0087] S32. By performing feature extraction and pattern matching on the input data, the current operating status of the drive shaft is determined.

[0088] S33. When a fault is detected in the drive shaft, the location of the fault is determined by analyzing the propagation path and weight distribution of each data acquisition point in the neural network.

[0089] In step S31, a multilayer perceptron, a common architecture in neural network models, is used to establish the framework for the fault diagnosis algorithm model. In driveshaft fault diagnosis, its input layer receives preprocessed sensor data feature vectors. Assuming that the operating parameters in an abnormal state have n features, the input layer has n neurons. Multiple hidden layers are typically set to learn complex patterns in the data. For example, three hidden layers are set: the first layer has 128 neurons, the second layer has 64, and the third layer has 32. The hidden layers are connected by weight matrices. The neurons perform weighted summation of the inputs and are processed by a nonlinear activation function, such as the ReLU function: f(x) = max(0, x), to increase the model's nonlinear expressive power. The number of neurons in the output layer is equal to the number of fault types, m. The softmax function is used as the activation function to convert the output value into a probability distribution, representing the probability that the driveshaft is in each fault type.

[0090] In step S32, the neural network structure within the fault diagnosis algorithm model processes the input data layer by layer, from the input layer to the hidden layer and then to the output layer. Each neuron performs weighted summation and nonlinear transformation on the data according to its weights to extract feature patterns from the data. The softmax function of the output layer converts the output value into a probability distribution. The category with the highest probability is the possible current operating state of the drive shaft predicted by the fault diagnosis algorithm model. If the category with the highest probability is the normal operating category and the probability exceeds a certain threshold, such as a threshold of 0.8, then the drive shaft is considered to be operating normally; otherwise, the drive shaft is considered to have a possible fault, and the fault type with the highest probability is output.

[0091] In step S33, when the fault diagnosis algorithm model determines that the drive shaft has a fault, it roughly determines the location of the fault by analyzing the data collected from each data acquisition point, i.e., the propagation path and weight distribution of the data collected by each sensor in the neural network. Each layer and each neuron in the neural network has specific connection weights with the input sensor data. These weights reflect the "degree of attention" that neuron pays to different sensor data features. For example, when a specific frequency abnormal peak related to the imbalance fault is detected, the fault diagnosis algorithm model traces the propagation path of the frequency signal in the neural network to see which neurons play a key role in processing the frequency feature and which sensor data are closely related to these neurons. By analyzing the weight distribution of these neurons, the contribution of different sensor data to identifying the fault feature can be understood. Assuming that the key neurons related to the specific frequency mainly receive data from vibration sensors near a certain end of the drive shaft, and that these sensor data have high weights in the neural network, then it can be preliminarily determined that the imbalance may occur at this specific part of the drive shaft.

[0092] Through the detailed fault diagnosis algorithm process described above, the fault diagnosis algorithm model based on deep learning neural networks can efficiently and accurately diagnose faults in the drive shaft of hydro-generator units, providing strong support for the stable operation of the equipment.

[0093] like Figure 11 As shown, in step S4, a multi-parameter fusion evaluation model is constructed to analyze the severity of the fault corresponding to the operating parameters in an abnormal state, including the following steps: S41. For each type of operating parameter related to drive shaft failure, select multiple specific parameter indicators to assess the severity of the failure.

[0094] S42. Determine the weight of each parameter index in explaining the severity of the fault through weight analysis.

[0095] S43. Multiply the actual value of each parameter index by its corresponding weight, and then add up the weighted values ​​of all parameter indexes to obtain a comprehensive fault severity index.

[0096] In step S41, in addition to vibration amplitude, vibration frequency, phase, and vibration modes are selected as parameters related to transmission faults to assess the severity of the fault. For example, different types of faults will produce significant vibration responses at specific frequencies. Imbalance faults typically show a significant increase in vibration amplitude at the first harmonic; while misalignment faults may lead to an increase in vibration components at the second harmonic or higher harmonics. Phase information can help determine the specific location and nature of the fault; for example, a phase abrupt change may indicate loosening or displacement of components. Vibration modal analysis can further reveal the overall vibration pattern of the drive shaft. Different faults will cause changes in vibration modes. By monitoring changes in vibration modes, a more comprehensive understanding of the fault situation can be obtained.

[0097] For temperature parameters related to transmission failures, in addition to real-time temperature values ​​and temperature change rates, the temperature field distribution must also be considered to assess the severity of the failure. Localized overheating may be caused by poor lubrication, component friction, or cooling system malfunctions. By analyzing the temperature field distribution, the location and extent of the overheated area can be determined, thereby assessing the severity of the failure. For example, in the bearing area of ​​a drive shaft, if there is an uneven temperature field, with local temperatures significantly higher than other areas, it may indicate problems such as bearing wear or improper installation. In this case, the severity of the failure is relatively high.

[0098] For stress parameters related to transmission failures, stress magnitude, stress variation range, and stress concentration factor are important parameters for assessing the severity of transmission shaft failures. The stress concentration factor reflects the degree of stress concentration in a localized area of ​​a component; a high stress concentration factor indicates that the area is more prone to fatigue cracking and damage. An increased stress variation range may indicate unstable loads on the transmission bearing, accelerating fatigue damage to the component. For example, stress concentration is prone to occur in keyways and shoulders of transmission shafts due to structural discontinuities. By monitoring the stress concentration factor and stress variation range in these areas, potential faults can be detected in a timely manner, and their severity can be assessed.

[0099] In step S42, the analytic hierarchy process (AHP) can be used to decompose multiple stress parameters related to transmission faults into multiple levels. The weights are determined by comparing the relative importance of elements at each level. In assessing the severity of transmission shaft faults, a hierarchical model is first constructed. The target level is set as the fault severity assessment level, the criterion level includes parameter categories such as vibration, temperature, and stress, and the scheme level contains specific parameter indicators. Then, the relative importance of elements at each level is determined through expert experience or pairwise comparison matrices, and the weight of each parameter indicator is calculated. For example, for imbalance faults, experts believe, based on experience, that vibration parameters have the greatest impact on fault severity assessment, followed by temperature parameters, while stress parameters have a relatively smaller impact. The weights of specific parameters such as vibration amplitude, frequency, temperature, and stress are calculated using the AHP.

[0100] In step S43, a weighted summation method is used. The actual value of each parameter is multiplied by its corresponding weight, and then the weighted values ​​of all parameters are summed to obtain a comprehensive fault severity index. For example, let the vibration amplitude weight be w1, and the actual value be V1; the temperature change rate weight be w2, and the actual value be T2; and the stress concentration factor weight be w3, and the actual value be S3. Then the comprehensive fault severity index I = w1V1 + w2T2 + w3S3. Based on this index, the severity of the fault can be divided into different levels, such as minor, moderate, and severe, and a corresponding threshold range can be set for each level.

[0101] like Figure 12 As shown, in step S5, a corresponding maintenance plan is generated based on the fault type and severity, including the following steps: S51. For minor driveshaft imbalance issues, a dynamic balancing test and correction can be performed on the driveshaft during the next scheduled maintenance. Detailed testing procedures and standards are provided. The maintenance plan lists the necessary tools, such as a dynamic balancing tester and wrenches. The operating procedures include how to install the sensors on the dynamic balancing tester, how to conduct the test, and how to adjust the counterweights based on the test results. Standards and acceptance criteria for the dynamic balancing test are also provided to ensure the maintenance work achieves the desired results.

[0102] S52. For more complex or rare faults, deep learning algorithms are used to analyze a large number of historical fault cases and solutions to generate targeted maintenance plans. For example, when encountering a new type of drive shaft fault whose characteristics are not entirely the same as previously recorded faults, detailed information about the fault, including sensor data, fault manifestations, and fault development trends, is input into a deep learning-based case reasoning model. The model searches for similar fault cases in massive amounts of historical data, considering not only the similarity of fault type and severity but also factors such as the environment in which the fault occurred and the equipment's operating conditions. By comparing, analyzing, and optimizing solutions for similar cases, a personalized maintenance plan is generated for the current complex fault.

[0103] S53. During the scheme generation process, the optimal maintenance strategy is found by taking into account maintenance costs, maintenance time, and factors affecting unit operation.

[0104] Maintenance costs include direct and indirect costs. Direct costs include the cost of replacing parts, purchasing repair tools, and labor costs. Indirect costs include power generation losses due to maintenance downtime and the impact on grid stability. Costs are estimated by comparing different maintenance options. For example, when choosing to replace a component of a drive shaft, the prices and lifespans of different brands and quality levels are compared. Assuming brand A components are more expensive but have a longer lifespan, while brand B components are cheaper but have a relatively shorter lifespan, a cost-benefit analysis is conducted based on factors such as the equipment's expected operating time, maintenance cycles, and potential future replacement costs to select the component with the highest cost-effectiveness, minimizing maintenance costs while still achieving the desired maintenance results.

[0105] Maintenance time is also a crucial consideration, as prolonged downtime for maintenance can impact the power generation of hydroelectric generators and the stability of the power grid. A reasonable maintenance time plan should be developed based on the severity of the fault, the complexity of the maintenance task, and the availability of necessary resources. For example, for severe faults requiring rapid restoration of unit operation, experienced technicians should be prioritized to form a repair team, and necessary spare parts and tools should be allocated to complete the maintenance work in the shortest possible time. For maintenance tasks that can be performed during periods of unit operation, maintenance time should be rationally scheduled based on forecasts of low grid load periods to minimize the impact on power generation. By optimizing maintenance time, maintenance quality can be ensured while minimizing disruption to normal unit operation.

[0106] In addition to considering cost and time, the maintenance plan also takes into account other impacts on unit operation. For example, some maintenance operations may require the disassembly or adjustment of parts of the unit's systems, which could affect other components. By assessing the impact of different maintenance plans on the overall operation of the unit, the plan with the least impact on other components is selected. Furthermore, the maintenance plan includes corresponding preventative measures and follow-up monitoring recommendations to ensure stable unit operation during and after maintenance, avoiding new malfunctions caused by the maintenance operations.

[0107] Through the detailed intelligent maintenance solution generation process described above, the system can provide scientific, reasonable, and personalized maintenance solutions for various faults in the drive shaft of the hydro-generator unit, effectively improving the maintenance efficiency and operational reliability of the unit, and reducing maintenance costs and failure risks.

[0108] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0109] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. The above are merely embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made based on the description and drawings of this application, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

[0110] The specific embodiments of the invention have been described in detail above, but they are only examples, and this application is not limited to the specific embodiments described above. For those skilled in the art, any equivalent modifications or substitutions to the invention are also within the scope of this application. Therefore, all equivalent changes, modifications, and improvements made without departing from the spirit and principles of this application should be covered within the scope of this application.

Claims

1. A method for fault maintenance of the rotating shaft of a mechanical oil pump in a hydro-generator unit based on multi-parameter fusion, characterized in that, Includes the following steps: The operating parameters of the rotating shaft during operation are acquired in real time, including temperature parameters, vibration parameters, and stress parameters, as well as the operating load and speed of the hydro-generator unit corresponding to the operating parameters. Construct a normal operating parameter model for the drive shaft, and determine whether the operating parameters are within the normal range based on the normal operating parameter model; Construct a fault diagnosis algorithm model, and analyze the fault type corresponding to the operating parameters in an abnormal state based on the fault diagnosis algorithm model; Construct a multi-parameter fusion evaluation model to analyze the severity of faults corresponding to operating parameters in abnormal states; A corresponding maintenance plan is generated based on the fault type and severity.

2. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 1, characterized in that, The operating parameters of the rotating shaft during operation are acquired in real time, including the following steps: Triaxial vibration sensors are installed at the bearing seats near both ends of the drive shaft to accurately measure vibration parameters along the X, Y, and Z directions, including vibration amplitude, vibration frequency, and vibration phase. A surface temperature sensor is used on the drive shaft to collect the surface temperature of the drive shaft in real time. Strain gauge stress sensors are installed at the connection points of the coupling. By detecting the minute strain generated when the coupling is subjected to force, the stress is converted into an electrical signal to measure the magnitude of the stress.

3. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 1, characterized in that, Constructing a normal operating parameter model for the drive shaft includes the following steps: Normal operating parameters of the drive shaft under different working conditions are collected under normal operating conditions, including normal temperature parameters, normal vibration parameters and normal stress parameters under different load and speed conditions. The collected normal operation parameters are processed for time synchronization. The collected normal operation parameters are subjected to feature extraction and transformation to mine representative and discriminative feature variables; The feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions; The parameter models under different operating conditions are integrated to form a complete normal operation parameter model library.

4. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 3, characterized in that, The collected normal operation parameters are subjected to feature extraction and transformation to mine representative and discriminative feature variables, including the following steps: To mine the characteristic variables of normal temperature parameters, the following steps are performed: Extract basic temperature features, including the current temperature value, the rate of temperature change, and the temperature mean and variance; Extract temperature trend features, including long-term temperature trend features and seasonal temperature change features; For the mining of characteristic variables of normal vibration parameters, the following steps are performed: Extract the time-domain features of normal vibration parameters, including vibration amplitude, peak factor, and kurtosis; Extract the frequency domain characteristics of normal vibration parameters, including spectral analysis features, power spectral density, and frequency centroid; For characteristic variable mining of normal stress parameters, the following steps are performed: Extract stress magnitude characteristics, including stress amplitude, average stress, and stress ratio; Extract stress cycle characteristics, including the number of stress cycles and cyclic stress characteristic parameters.

5. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 3 or 4, characterized in that, The process of extracting and transforming features from the collected normal operation parameters to uncover representative and discriminative feature variables also includes the following steps: Extract load-related features, including the correlation between load and normal vibration parameters, and the relationship between load and normal temperature parameters; Extract rotational speed-related features, including the matching between rotational speed and normal vibration frequency, and the relationship between rotational speed and normal temperature and stress parameters.

6. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 3, characterized in that, The feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions, including the following steps: The feature variables are organized into a dataset in a unified format. Each data record in the dataset contains information on multiple dimensions of the drive shaft at a certain moment, including vibration amplitude, frequency characteristics, temperature value, temperature change rate, stress magnitude, and stress ratio. Based on the characteristics of data distribution and the needs of practical applications, a suitable clustering algorithm is determined to perform clustering analysis on the data in the dataset. The quality of clustering is evaluated by calculating the silhouette coefficient and CH index of the clustering results. Perform statistical analysis on each clustering result and calculate the statistical parameters of the mean, median, standard deviation, minimum, and maximum of each feature variable; Based on the statistical parameters of each clustering result, the normal operating parameter range of the drive shaft under different working conditions is determined.

7. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 6, characterized in that, The feature variables are input into the optimized clustering algorithm for cluster analysis to obtain parameter models under different working conditions. The process also includes the following steps: By calculating the correlation coefficients between various feature variables, the linear or nonlinear correlation between different feature variables can be determined. Based on the results of feature correlation analysis, a feature relationship model is constructed among the feature variables; By integrating the normal operating parameter ranges and characteristic relationship models under different working conditions, a complete normal operating parameter model for the drive shaft is formed.

8. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 1, characterized in that, Construct a fault diagnosis algorithm model, and analyze the fault type corresponding to the operating parameters in an abnormal state based on the fault diagnosis algorithm model, including the following steps: A fault diagnosis algorithm model is established based on a neural network architecture, and the operating parameters in an abnormal state are input into the fault diagnosis algorithm model. The current operating status of the drive shaft is determined by performing feature extraction and pattern matching on the input data. When a fault is detected in the drive shaft, the location of the fault is determined by analyzing the propagation path and weight distribution of each data acquisition point in the neural network.

9. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 1, characterized in that, Construct a multi-parameter fusion evaluation model to analyze the severity of faults corresponding to operating parameters in abnormal states, including the following steps: For each type of operating parameter related to drive shaft failure, multiple specific parameter indicators are selected to assess the severity of the failure. The weight of each parameter index in explaining the severity of the fault was determined by weight analysis. Multiply the actual value of each parameter by its corresponding weight, and then sum the weighted values ​​of all parameters to obtain a comprehensive fault severity index.

10. The method for maintaining the mechanical oil pump rotating shaft of a hydro-generator unit based on multi-parameter fusion according to claim 1, characterized in that, Based on the fault type and severity, a corresponding maintenance plan is generated, including the following steps: For minor driveshaft imbalance faults, dynamic balancing tests and corrections can be performed on the driveshaft during the next scheduled maintenance, and detailed test procedures and standards will be provided. For more complex or rare faults, deep learning algorithms are combined to analyze a large number of historical fault cases and solutions to generate targeted maintenance plans. During the scheme generation process, the optimal maintenance strategy is found by taking into account maintenance costs, maintenance time, and factors affecting unit operation.

Images