Axis measuring system for hydroelectric generating set

Abstract

The invention provides a hydroelectric generating set axis measuring system, relates to the field of hydroelectric generating set management, and adopts a data analysis technology based on artificial intelligence. By analyzing the X-direction and Y-direction displacement values of an upper guide bearing, the X-direction and Y-direction displacement values of a lower guide bearing, the X-direction and Y-direction displacement values of a flange and the X-direction and Y-direction displacement values of a hydroelectric generating set, which are measured at multiple angles in the automatic barring rotation process of the hydroelectric generating set, whether the axis of the hydroelectric generating set is bent or not is judged. In this way, the state of the axis of the hydroelectric generating set can be reflected more comprehensively through the displacement data of the multiple angles, and then the accuracy and reliability of tortuosity judgment can be improved.

Description

Technical Field

[0001] This application relates to the field of hydropower unit management, and more specifically, to a hydropower unit axis measurement system. Background Art

[0002] The main shaft of a vertical hydro-generator set consists of the main shaft of the generator and the main shaft of the water turbine. The two are connected by end flanges and bolts, and the center line passing through the main shaft of the unit is used to represent the actual axis of the unit. In theory, after the generator shaft and the water turbine shaft are connected, the axis of the unit formed should be a straight line. However, in actual applications, due to manufacturing and installation errors, the main axis of the unit does not completely coincide with its rotation center line, and there are varying degrees of inclination or twists. Such twists will cause the axis of the unit to deviate from the theoretical rotation center during operation, resulting in increased vibration and swing amplitude of the unit, which will pose a threat to the safe operation of the unit and may even directly cause damage to the unit.

[0003] Therefore, a hydropower unit axis measurement solution is needed. Summary of the Invention

[0004] This application aims at the deficiencies in the prior art and provides a hydropower unit axis measurement system.

[0005] According to one aspect of this application, a hydropower unit axis measurement system is provided, which includes: A hydropower unit related data acquisition module for obtaining the X and Y displacement values of the upper guide bearing, the X and Y displacement values of the lower guide bearing, the X and Y displacement values of the flange, and the X and Y displacement values of the water guide measured by a laser sensor at multiple angles during the automatic barring rotation of the hydropower unit; A hydropower unit related data sorting module for structurally regularizing the X and Y displacement values of the upper guide bearing, the X and Y displacement values of the lower guide bearing, the X and Y displacement values of the flange, and the X and Y displacement values of the water guide measured at the multiple angles to obtain a unit axis measurement displacement input matrix; A hydropower unit related data processing module for encoding the unit axis measurement displacement input matrix to obtain an optimized unit axis measurement displacement fusion feature vector; A unit axis twist state result generation module for judging the twist state of the hydropower unit axis based on the optimized unit axis measurement displacement fusion feature vector.

[0006] In the above hydropower unit axis measurement system, the hydropower unit related data sorting module is used to: arrange the X and Y displacement values of the upper guide bearing, the X and Y displacement values of the lower guide bearing, the X and Y displacement values of the flange, and the X and Y displacement values of the water guide measured at the multiple angles according to the angle dimension to obtain the unit axis measurement displacement input matrix.

[0007] In the above-mentioned hydropower unit axis measurement system, the hydropower unit related data processing module includes: a unit axis measurement displacement feature extraction unit, which is used to perform preliminary feature extraction on the unit axis measurement displacement input matrix to obtain a multi-scale unit axis measurement displacement feature matrix; a unit axis measurement displacement feature emphasis unit, which is used to perform feature emphasis on the multi-scale unit axis measurement displacement feature matrix to obtain a unit axis measurement displacement emphasis feature matrix; a unit axis measurement displacement fusion feature generation unit, which is used to fuse the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain the optimized unit axis measurement displacement fusion feature vector.

[0008] In the above-mentioned hydropower unit axis measurement system, the unit axis measurement displacement feature extraction unit is used to: pass the unit axis measurement displacement input matrix through the multi-scale axis measurement displacement feature extractor to obtain the multi-scale unit axis measurement displacement feature matrix.

[0009] In the above-mentioned hydropower unit axis measurement system, the unit axis measurement displacement feature emphasis unit is used to: pass the multi-scale unit axis measurement displacement feature matrix through the axis measurement displacement feature emphasizer to obtain the unit axis measurement displacement emphasis feature matrix.

[0010] In the above-mentioned hydropower unit axis measurement system, the axis measurement displacement feature emphasizer is a convolutional neural network model using a spatial attention mechanism, and the multi-scale axis measurement displacement feature extractor is a convolutional neural network model including a first convolutional layer and a second convolutional layer, wherein the first convolutional layer uses a two-dimensional convolutional kernel of a first scale, and the second convolutional layer uses a two-dimensional convolutional kernel of a second scale, and the first scale is different from the second scale.

[0011] In the above-mentioned hydropower unit axis measurement system, the unit axis measurement displacement fusion feature generation unit includes: a unit axis measurement displacement feature matrix dimension reduction subunit, which is used to expand the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain a multi-scale unit axis measurement displacement feature vector and a unit axis measurement displacement emphasis feature vector; a unit axis measurement displacement fusion feature generation subunit, which is used to perform weighted fusion on the multi-scale unit axis measurement displacement feature vector and the unit axis measurement displacement emphasis feature vector to obtain a unit axis measurement displacement fusion feature vector; a unit axis measurement displacement fusion feature optimization subunit, which is used to perform feature kernel domain projection modulation based on intrinsic regression on the unit axis measurement displacement fusion feature vector to obtain an optimized unit axis measurement displacement fusion feature vector.

[0012] In the above-mentioned hydropower unit axis measurement system, the unit axis measurement displacement fusion feature optimization subunit is used to: construct a pixel-by-pixel granularity information association matrix of the unit axis measurement displacement fusion feature vector; perform kernel feature extraction on the pixel-by-pixel granularity information association matrix based on the convolution layer to obtain a unit axis measurement displacement fusion feature kernel domain nonlinear activation matrix; perform feature spectrum decomposition on the pixel-by-pixel granularity information association matrix to obtain a set of unit axis measurement displacement fusion feature intrinsic component coding vectors; and convert each unit axis measurement displacement fusion feature intrinsic component coding vector in the set of unit axis measurement displacement fusion feature intrinsic component coding vectors into a nonlinear activation matrix. Input a feature significant modulation unit based on the self-attention mechanism to obtain a set of significant modulation coding vectors of the unit axis measurement displacement fusion feature intrinsic component; project each unit axis measurement displacement fusion feature intrinsic component significant modulation coding vector in the set of the unit axis measurement displacement fusion feature intrinsic component significant modulation coding vectors to the unit axis measurement displacement fusion feature kernel domain nonlinear activation matrix to obtain a set of unit axis measurement displacement fusion feature intrinsic component kernel mask coding vectors; fuse the set of unit axis measurement displacement fusion feature intrinsic component kernel mask coding vectors to obtain the optimized unit axis measurement displacement fusion feature vector.

[0013] In the above-mentioned hydropower unit axis measurement system, the unit axis tortuosity state result generation module is used to: pass the optimized unit axis measurement displacement fusion feature vector through the unit axis tortuosity state classifier to obtain a tortuosity state classification result, and the tortuosity state classification result is used to indicate whether the hydropower unit axis is tortuous.

[0014] This application has significant technical effects due to the adoption of the above technical solutions: The hydroelectric unit axis measurement system provided by the present application adopts data analysis technology based on artificial intelligence, and determines whether the axis of the hydroelectric unit is tortuous by analyzing the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at multiple angles during the automatic turning process of the hydroelectric unit. In this way, the displacement data at multiple angles can more comprehensively reflect the state of the axis of the hydroelectric unit, which is conducive to improving the accuracy and reliability of the tortuosity judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 It is a system block diagram of a hydropower unit axis measurement system according to an embodiment of the present application.

[0016] Figure 2 Schematic diagram of data flow of a hydroelectric generator unit axis measurement system according to an embodiment of the present application.

[0017] Figure 3 It is a block diagram of a hydropower unit related data processing module in a hydropower unit axis measurement system according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0019] In the field of hydropower generation, vertical turbine generator sets play a vital role. Their main shaft structure consists of the generator main shaft and the turbine main shaft. When designing and assembling, these two main shafts are tightly connected through the flanges at the ends and with the help of high-strength bolts to ensure that the entire main shaft can work together, transmit torque, and then drive the generator rotor to rotate stably, realizing efficient conversion of water energy to electrical energy.

[0020] Under ideal theoretical conditions, when the generator shaft and the turbine shaft are connected, the axis of the unit should be a straight line without any deviation. However, in actual engineering application scenarios, on the one hand, due to the limitations of the processing technology and the subtle differences in the production process of various parts, it is difficult for the dimensional accuracy, form and position tolerances of both the generator main shaft and the turbine main shaft to be completely consistent with the design theoretical values; on the other hand, the environmental conditions at the installation site, the operating level of the construction personnel, and the accuracy of the installation tools and equipment used will all have an effect on the final installation effect. In this way, the main axis of the unit will not completely coincide with its rotation center line, and there will be different degrees of inclination or twists, which will cause the axis of the unit to deviate from the theoretical rotation center during operation, thereby causing the unit to vibrate more and swing more, which will pose a threat to the safe operation of the unit and may even directly cause damage to the unit. Therefore, a hydropower unit axis measurement solution is expected.

[0021] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for the measurement of the axis of hydropower units.

[0022] Figure 1 It is a system block diagram of a hydropower unit axis measurement system according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the hydroelectric generator unit axis measurement system according to an embodiment of the present application. Figure 1 and Figure 2 As shown, in the hydropower unit axis measurement system 100, it includes: a hydropower unit related data acquisition module 110, which is used to obtain the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured by the laser sensor at multiple angles during the automatic turning of the hydropower unit; a hydropower unit related data sorting module 120, which is used to structure and regularize the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at the multiple angles to obtain a unit axis measurement displacement input matrix; a hydropower unit related data processing module 130, which is used to encode the unit axis measurement displacement input matrix to obtain an optimized unit axis measurement displacement fusion feature vector; a unit axis tortuosity state result generation module 140, which is used to judge the tortuosity state of the hydropower unit axis based on the optimized unit axis measurement displacement fusion feature vector.

[0023] In the embodiment of the present application, the hydroelectric unit related data acquisition module 110 is used to obtain the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured by the laser sensor at multiple angles during the automatic turning of the hydroelectric unit. It should be understood that the upper guide bearing is located at the upper part of the generator main shaft, mainly used to support the main shaft and ensure its stability. The lower guide bearing is located at the lower part of the generator main shaft, and its function is similar to that of the upper guide bearing, providing additional support and stability. The flange is a component that connects the generator main shaft and the turbine main shaft, usually connected by bolts. The docking accuracy of the flange directly affects the degree of centering of the main shaft, so its position and state are crucial to the overall performance of the unit. The water guide is a component inside the turbine, responsible for guiding the water flow to improve the efficiency of the turbine. The four positions of the upper guide bearing, the lower guide bearing, the flange, and the water guide are all key structural components in the hydroelectric unit, and their displacement conditions can directly reflect the state of the unit axis. By measuring the X-direction (horizontal direction) and Y-direction (vertical direction) displacement values ​​of these positions, we can fully understand the deformation and offset information of the unit during the rotation process. That is, by comparing the displacement values ​​of these four positions, we can analyze the offset of the unit axis in different directions. If the displacement value of a certain position is abnormally large or small, or there is a significant difference in the displacement values ​​between different positions, it may mean that there are problems such as tortuosity or tilt in the axis of the unit. Therefore, these displacement values ​​are an important basis for judging the tortuosity state of the axis of the hydropower unit. It is also considered that the hydropower unit may be affected by various factors during operation, resulting in different degrees of offset of the axis of the unit in different directions. In order to understand the state of the axis of the unit more comprehensively and accurately, it is necessary to measure and analyze from multiple angles to capture the displacement changes of the unit in different directions, so as to more accurately judge the tortuosity state of the axis.

[0024] In order to accurately determine whether the axis of the hydropower unit is tortuous, obtaining the displacement values ​​of multiple key parts at multiple angles during the automatic turning process is a key basic work. The following will elaborate on the implementation content of this process in reality.

[0025] First, during the equipment installation and preparation stage, laser sensors need to be carefully selected and installed. For the upper guide bearing, since it is located on the upper part of the generator main shaft and mainly supports the main shaft and ensures stability, the sensor should be installed in a position that can accurately capture its displacement changes in the X direction (horizontal direction) and Y direction (vertical direction). Usually, according to the structural characteristics of the upper guide bearing, a suitable mounting bracket or fixture is selected to ensure that the sensor is firm and the measurement direction is accurate. During the installation process, the relevant technical specifications must be strictly followed to ensure that the relative position relationship between the sensor and the upper guide bearing is stable and not affected by factors such as unit operation vibration.

[0026] The lower guide bearing is located at the bottom of the generator main shaft. Its function is similar to that of the upper guide bearing. It also needs to be installed with a laser sensor to measure the displacement value. During installation, in addition to considering the accuracy of the measurement direction, it is also necessary to consider the compatibility with other components and measuring equipment to avoid mutual interference. After the sensor is installed, it needs to be accurately calibrated. The calibration work requires the use of professional calibration equipment or reference objects with known precise displacement. By comparing the sensor measurement value with the standard value, the internal parameters of the sensor, such as gain, zero offset, etc., are adjusted to control the measurement error within a very small range to ensure the high accuracy and reliability of the measurement data. The calibration process must be carried out in strict accordance with the sensor's operating manual, and the calibration data must be recorded in detail. These data will serve as an important reference in subsequent data analysis to evaluate the credibility of the measurement results.

[0027] The flange is a key component that connects the main shaft of the generator and the main shaft of the turbine. Its position and state have a significant impact on the overall performance of the unit. The laser sensor installed on the flange should be aligned with the key measurement points of the flange, which are usually selected at positions that can reflect the overall displacement trend and deformation of the flange. The installation of the sensor should ensure that it will not loosen or deviate due to factors such as vibration and temperature changes during the operation of the unit, especially when the automatic turning gear is rotating, so as to ensure the continuity and accuracy of the measurement data.

[0028] The water guide is a component inside the turbine that guides the water flow to improve the efficiency of the turbine. Its displacement can also reflect the state of the unit axis. When installing the laser sensor at the water guide, the complex working environment inside the turbine, such as water flow impact, high humidity and other factors, must be fully considered. The sensor should have good waterproof and moisture-proof properties, and the installation position should be able to accurately measure the displacement value of the water guide in the X and Y directions without hindering the normal water flow guiding function of the turbine.

[0029] After completing the installation and calibration of the sensors, all laser sensors need to be connected to the data acquisition system. During the connection process, ensure that the line connection is firm, the contact is good, and the signal transmission is stable and reliable. The selection of the data acquisition system is also crucial. It should have high-precision data acquisition capabilities, sufficient storage capacity, and flexible data processing functions. After the connection is completed, the data acquisition system is fully debugged. Set a suitable sampling frequency. The selection of the sampling frequency should take into account factors such as the unit rotation speed, measurement accuracy requirements, and data storage and processing capabilities. If the sampling frequency is too high, it may cause too much data, which will bring difficulties to data storage and processing; if the sampling frequency is too low, some key displacement change information may be missed, affecting the accuracy of the axis state judgment. Generally speaking, according to the specific operating characteristics and measurement requirements of the unit, the sampling frequency can be set from several times to dozens of times per second. At the same time, the data storage format must be set to ensure that the data can be easily analyzed and processed later. During the debugging process, a pre-acquisition test is performed to check whether the data acquisition system can accurately receive and record the displacement data from each sensor, including data integrity, accuracy, and time synchronization.

[0030] When the automatic cranking device is started, the main shaft of the unit will rotate slowly at the predetermined speed and direction. Before starting the automatic cranking, ensure that all components of the unit are in normal condition, the lubrication system is working well, the cranking device itself is running smoothly, and there is no jamming, abnormal noise, etc. The cranking speed should be selected moderately, ensuring that enough measurement data can be obtained to fully reflect the state of the unit axis, but not too fast so that the sensor cannot accurately capture the displacement change or the measurement accuracy is affected by factors such as centrifugal force. In general, the cranking speed can be controlled between a few revolutions per minute and dozens of revolutions per minute. The specific speed can be determined according to factors such as the unit model, size, and measurement requirements.

[0031] During the automatic turning process, the data acquisition program of the laser sensor must be strictly synchronized with the turning operation. The sensor measures the X- and Y-direction displacement values ​​of the upper guide bearing, lower guide bearing, flange and water guide at various angles in real time according to the preset sampling frequency. The data acquisition system records and stores these measurement data in a timely manner to ensure that the displacement data at each angle can accurately correspond to the corresponding measurement time and turning position. In order to achieve accurate angle measurement and data correspondence, an angle encoder or other angle measurement equipment is usually installed on the turning device or the unit main shaft to synchronize the angle position information of the turning with the displacement data collected by the sensor. In this way, in the subsequent data analysis, the displacement change law of the key parts of the unit under different position states can be accurately analyzed based on the angle position information.

[0032] During the turning process of the unit for alignment check, it is necessary to ensure that the laser sensor can cover the measurement of multiple angles. This requires reasonable planning of the turning angle range of the alignment check and the measurement angle range of the sensor. Generally, the turning of the alignment check is set to 360 degrees or its specific angle range to comprehensively obtain the displacement change of the unit within a complete turning cycle. The sensor collects data at each certain angle interval (such as 1 degree, 5 degrees, etc.). A smaller angle interval can obtain more detailed displacement change information, but it will also increase the amount of data. In actual operation, it is necessary to weigh and select a suitable angle interval according to the specific situation of the unit and the measurement accuracy requirements. For example, for large hydropower units or units with high requirements for the axis state, a smaller angle interval can be selected for measurement; while for some small units or preliminary detection situations, the angle interval can be appropriately increased. In this way, the displacement measurement of the key parts of the unit in different position states can be realized, and the change of the axis state of the unit during the turning process can be comprehensively reflected.

[0033] During the data acquisition process, the collected data should be monitored and preliminarily processed in real time. The data acquisition system should have the functions of data display and preliminary analysis. The operator can observe in real time whether the collected displacement data is within a reasonable range and whether there are obvious abnormal fluctuations or jumps. If data anomalies are found, the working state of the sensor, the connection lines, and the operation of the unit should be checked in time to eliminate faults or interference factors. At the same time, some preliminary processing can be carried out on the collected data, such as data filtering. Since there may be various interference factors in the measurement environment, such as electromagnetic interference and mechanical vibration, these interferences may cause noise in the measurement data. By adopting appropriate data filtering algorithms, such as moving average filtering and low-pass filtering, the high-frequency noise in the data can be removed, the data curve can be smoothed, and the change trend of the displacement value can be made clearer, which is convenient for subsequent data analysis and processing. However, in the filtering process, it is necessary to pay attention to selecting appropriate filtering parameters to avoid over-filtering resulting in data distortion and losing some important displacement change information.

[0034] After the data acquisition is completed, the recorded data needs to be further sorted out and processed. First, perform an integrity check on the data to ensure that there are complete displacement data records for each measurement point at various angles. If data is found to be missing, analyze the reasons, which may be caused by problems such as sensor failures or data transmission interruptions. For a small amount of missing data, reasonable interpolation can be made based on the data change trend and the data of adjacent measurement points; for a large amount of missing or abnormal data, re-measurement may be required. Second, verify the accuracy of the data by comparing and analyzing the collected data with the design parameters of the unit, historical operation data, and other relevant standards. If it is found that the displacement values of certain measurement points deviate significantly from the normal range, further check whether there are problems in the measurement process, such as whether the sensor calibration has failed or whether the installation position has changed. For abnormal data, detailed records and analysis should be carried out to determine its impact on the judgment of the axis state. If the abnormal data may affect the accuracy of the final result, corresponding measures should be considered to correct or re-measure.

[0035] After completing the sorting and preliminary processing of the data, the data needs to be stored and backed up. When storing data, select appropriate storage media and storage structures to ensure that the data can be stored for a long time and is convenient for retrieval and call. At the same time, to prevent data loss, regular data backups should be made, and various backup methods can be used, such as local backup, off-site backup, cloud backup, etc. The storage media of the backup data should also be regularly checked and maintained to ensure the integrity and availability of the backup data.

[0036] In summary, obtaining the displacement values of multiple key parts of a hydropower unit at multiple angles during the automatic turning process is a complex and rigorous process. From equipment installation and preparation, synchronization of automatic turning operation and data acquisition, to data recording and preliminary processing, and finally data storage and backup, each link needs to be carried out strictly in accordance with relevant technical specifications and operation procedures to ensure that the obtained data is accurate, reliable, and complete, providing a solid data basis for accurately judging the axis state of the hydropower unit in the future.

[0037] In the embodiment of the present application, the hydropower unit related data sorting module 120 is used to structure and regularize the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at the multiple angles to obtain the unit axis measurement displacement input matrix. Specifically, in the embodiment of the present application, the hydropower unit related data sorting module is used to: arrange the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at the multiple angles according to the angle dimension to obtain the unit axis measurement displacement input matrix. Accordingly, considering that the displacement value data at different angles and different positions measured by the laser sensor are relatively scattered, in order to organize these discrete data into an orderly and standardized form, it is convenient for subsequent data sorting and analysis, it is necessary to structure these displacement data. In order to reflect the displacement of the hydropower unit at different angles through these data, and then to extract characteristic information that can reflect the state of the unit axis, in the technical solution of the present application, it is first necessary to arrange the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at the multiple angles according to the angle dimension. In the arranged matrix, each row of data represents the displacement of the unit at a certain specific angle. That is, each row of data contains the X-direction and Y-direction displacement values ​​of the upper guide bearing, the lower guide bearing, the flange and the water guide at this angle. These displacement values ​​can reflect the deformation and offset of the unit at this angle. And each column of data represents the displacement of a certain position of the unit at different angles. This structured data form helps to improve the efficiency and accuracy of the analysis, thereby better judging the state of the axis of the hydropower unit.

[0038] In the embodiment of the present application, the hydropower unit related data processing module 130 is used to encode the unit axis measurement displacement input matrix to obtain an optimized unit axis measurement displacement fusion feature vector. Specifically, Figure 3 FIG. 1 is a block diagram of a hydropower unit-related data processing module in a hydropower unit axis measurement system according to an embodiment of the present application. Figure 3As shown, the hydropower unit related data processing module 130 includes: a unit axis measurement displacement feature extraction unit 131, used for performing preliminary feature extraction on the unit axis measurement displacement input matrix to obtain a multi-scale unit axis measurement displacement feature matrix; a unit axis measurement displacement feature emphasis unit 132, used for performing feature emphasis on the multi-scale unit axis measurement displacement feature matrix to obtain a unit axis measurement displacement emphasis feature matrix; a unit axis measurement displacement fusion feature generation unit 133, used for fusing the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain the optimized unit axis measurement displacement fusion feature vector.

[0039] In the embodiment of the present application, the unit axis measurement displacement feature extraction unit 131 is used to perform preliminary feature extraction on the unit axis measurement displacement input matrix to obtain a multi-scale unit axis measurement displacement feature matrix. Specifically, in the embodiment of the present application, the unit axis measurement displacement feature extraction unit is used to: pass the unit axis measurement displacement input matrix through a multi-scale axis measurement displacement feature extractor to obtain the multi-scale unit axis measurement displacement feature matrix. It should be understood that in order to identify the displacement modes at different positions at different angles from the unit axis measurement displacement input matrix, and then help determine the state of the hydropower unit axis, it is necessary to perform feature extraction processing on the unit axis measurement displacement input matrix. If the hydropower mechanism axis is tortuous, this will lead to nonlinear changes in displacement. Feature extraction can help capture this nonlinear feature, thereby identifying signs of tortuosity in the data. And by extracting effective features, the influence of noise and redundant information in the original matrix can be reduced, thereby improving the accuracy of judging the tortuosity state. In the technical solution of the present application, the unit axis measurement displacement input matrix is ​​input into the multi-scale axis measurement displacement feature extractor to capture relevant features. In the present application, the multi-scale axis measurement displacement feature extractor specifically refers to a convolutional neural network model including a first convolutional layer and a second convolutional layer, wherein the first convolutional layer uses a two-dimensional convolutional kernel of a first scale, and the second convolutional layer uses a two-dimensional convolutional kernel of a second scale, and the first scale is different from the second scale. The two-dimensional convolutional kernel of the first scale usually has a smaller receptive field and can capture subtle displacement changes. This convolutional kernel can help identify local features, such as displacement anomalies in a small range; while the two-dimensional convolutional kernel of the second scale has a larger receptive field and can capture features in a larger range. This convolutional kernel can help identify global features, such as overall displacement trends and patterns. By using convolutional kernels of different scales, local and global features can be captured at the same time. This multi-scale processing method enables the model to understand the displacement state of the hydropower unit more comprehensively. And this multi-scale feature extraction can enhance the robustness of the model to noise and interference, making feature extraction more stable.

[0040] In an embodiment of the present application, the unit axis measurement displacement feature emphasis unit 132 is used to perform feature emphasis on the multi-scale unit axis measurement displacement feature matrix to obtain the unit axis measurement displacement emphasized feature matrix. Specifically, in an embodiment of the present application, the unit axis measurement displacement feature emphasis unit is used to: pass the multi-scale unit axis measurement displacement feature matrix through the axis measurement displacement feature emphasizer to obtain the unit axis measurement displacement emphasized feature matrix. It should be understood that in the preliminary feature extraction process, a large number of features may be generated, and some of them are more critical to judging the tortuosity of the unit. In order to highlight these important features and thus improve the effectiveness of subsequent analysis, it is necessary to perform feature emphasis processing on the original multi-scale unit axis measurement displacement feature matrix. And feature emphasis can suppress the influence of noise or irrelevant features in the feature extraction process, so that the model is more focused on useful information. In the technical solution of the present application, the multi-scale unit axis measurement displacement feature matrix is ​​input into the axis measurement displacement feature emphasizer for feature emphasis processing. In the present application, the axis measurement displacement feature emphasizer specifically refers to a convolutional neural network model using a spatial attention mechanism. Those skilled in the art should be aware that the spatial attention mechanism can adaptively select and emphasize important features based on the input data, rather than relying on a fixed feature selection method, which enables accurate identification of key features and thus improves the judgment of tortuosity.

[0041] In an embodiment of the present application, the unit axis measurement displacement fusion feature generation unit 133 is used to fuse the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain the optimized unit axis measurement displacement fusion feature vector. Accordingly, considering that the multi-scale unit axis measurement displacement feature matrix captures the displacement characteristics of the unit axis at different spatial scales, it can reflect the overall deformation of the unit axis and the detail changes at different scales; and the unit axis measurement displacement emphasis feature matrix further extracts and emphasizes the key information in the unit axis displacement through the spatial attention mechanism, such as the mutation point of the displacement, the abnormal area, etc., and these emphasized features help to more accurately judge the state of the unit axis. The two feature matrices have certain complementarity in content. The multi-scale feature matrix provides comprehensive displacement information, while the emphasis feature matrix highlights the key information. In order to make full use of the advantages of the two feature matrices and improve the accuracy and reliability of the judgment, in the technical solution of the present application, the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix are fused to obtain the optimized unit axis measurement displacement fusion feature vector.

[0042] Specifically, in an embodiment of the present application, the unit axis measurement displacement fusion feature generation unit includes: a unit axis measurement displacement feature matrix dimensionality reduction subunit, used to expand the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain a multi-scale unit axis measurement displacement feature vector and a unit axis measurement displacement emphasis feature vector; a unit axis measurement displacement fusion feature generation subunit, used to weightedly fuse the multi-scale unit axis measurement displacement feature vector and the unit axis measurement displacement emphasis feature vector to obtain a unit axis measurement displacement fusion feature vector; a unit axis measurement displacement fusion feature optimization subunit, used to perform feature kernel domain projection modulation based on intrinsic regression on the unit axis measurement displacement fusion feature vector to obtain an optimized unit axis measurement displacement fusion feature vector.

[0043] In particular, although the construction of the fusion feature vector of the unit axis measurement displacement integrates multi-scale features and emphasized features, there may be problems of ignoring or underutilizing the internal structural information of the features during the processing process. This phenomenon is mainly due to the fact that the methods used in the feature extraction and fusion stages of the existing technology may not fully consider the complex correlations and potential spatial structural information between different features. For example, when extracting multi-scale features, although the feature changes at different scales can be identified, these features are often based on the results of statistical or frequency domain analysis, and it is difficult to fully capture the spatial distribution characteristics of the original data and the deep-level dependencies between the features. Similarly, in the process of feature weighted fusion, the simple weighted summation method may not effectively integrate the subtle differences and internal connections that are crucial to judging the axis state, resulting in the loss or weakening of some key feature information. In addition, although the existing machine learning models have strong pattern recognition capabilities, they still face challenges in how to accurately parse and fully utilize the rich structural information contained in these complex fused feature vectors, which limits the accuracy and reliability of the model in assessing the axis state of the hydropower unit under complex working conditions. Therefore, the unit axis measurement displacement fusion feature vector is subjected to feature kernel domain projection modulation based on intrinsic regression to obtain an optimized unit axis measurement displacement fusion feature vector.

[0044] More specifically, in an embodiment of the present application, the unit axis measurement displacement fusion feature optimization subunit is used to: construct a pixel-by-pixel granularity information association matrix of the unit axis measurement displacement fusion feature vector; perform kernel feature extraction on the pixel-by-pixel granularity information association matrix based on a convolution layer to obtain a unit axis measurement displacement fusion feature kernel domain nonlinear activation matrix; perform feature spectrum decomposition on the pixel-by-pixel granularity information association matrix to obtain a set of unit axis measurement displacement fusion feature intrinsic component coding vectors; and convert each unit axis measurement displacement fusion feature intrinsic component coding vector in the set of unit axis measurement displacement fusion feature intrinsic component coding vectors into a single unit axis measurement displacement fusion feature vector. Input a feature significant modulation unit based on the self-attention mechanism to obtain a set of significant modulation coding vectors of the unit axis measurement displacement fusion feature intrinsic component; project each unit axis measurement displacement fusion feature intrinsic component significant modulation coding vector in the set of the unit axis measurement displacement fusion feature intrinsic component significant modulation coding vectors to the unit axis measurement displacement fusion feature kernel domain nonlinear activation matrix to obtain a set of unit axis measurement displacement fusion feature intrinsic component kernel mask coding vectors; fuse the set of unit axis measurement displacement fusion feature intrinsic component kernel mask coding vectors to obtain the optimized unit axis measurement displacement fusion feature vector.

[0045] In the embodiment of the present application, a pixel-by-pixel granularity information association matrix of the unit axis measurement displacement fusion feature vector is constructed, which can be expressed as follows:

[0046] in, represents the unit axis measurement displacement fusion feature vector, and They represent the first and The eigenvalues ​​at the positions, Indicates the calculation of Euclidean distance, Represents the pixel-by-pixel granular information association matrix The eigenvalue of the location.

[0047] That is, by constructing a pixel-by-pixel association network of the displacement fusion feature vectors of the unit axis measurement, the discrete displacement measurement values ​​are essentially mapped to a graph structure that describes the system topology, so that the originally scattered physical signals are converted into interpretable correlation intensity maps, thereby providing high-order features with physical semantics for subsequent analysis. In this way, by quantifying the redundancy and complementarity between the displacement fusion feature vectors of the unit axis measurement, noise interference can be filtered out, and the key patterns related to the axis tortuosity can be highlighted. The generated pixel-by-pixel granular information association matrix significantly improves the sensitivity and anti-interference ability of the axis state judgment, and avoids systematic misdiagnosis caused by misjudgment of local features.

[0048] In the embodiment of the present application, kernel feature extraction is performed on the pixel-by-pixel granularity information correlation matrix based on a convolutional layer to obtain a non-linear activation matrix of the kernel domain of the displacement fusion feature of the unit axis measurement, which can be expressed by the formula:

[0049] Wherein, represents the pixel-by-pixel granularity information correlation matrix, represents the convolutional layer, represents the non-linear activation matrix of the kernel domain of the displacement fusion feature of the unit axis measurement That is, kernel feature extraction is performed on the pixel-by-pixel granularity information correlation matrix through a convolutional layer to automatically analyze the deep-seated dependency relationships hidden between features, thereby converting the discrete statistical correlations in the pixel-by-pixel granularity information correlation matrix into interpretable and physically meaningful high-order feature expressions. Specifically, context information with non-linear semantics is generated in a data-driven manner, enabling the system to distinguish systematic correlation anomalies caused by axis bending from random noise or local interference, and then providing a more discriminative feature representation for subsequent projection modulation. The generated non-linear activation matrix of the kernel domain of the displacement fusion feature of the unit axis measurement significantly improves the model's characterization ability of the axis bending state.

[0050] In the embodiment of the present application, spectral decomposition is performed on the pixel-by-pixel granularity information correlation matrix to obtain a set of eigen-component coding vectors of the displacement fusion feature of the unit axis measurement, which can be expressed by the formula:

[0051] Wherein, represents the transpose of the vector, represents a diagonal matrix, and respectively represent the first and the th eigenvalues on the diagonal of the diagonal matrix, represents a set of eigen-component coding vectors of the displacement fusion feature of the unit axis measurement, , , respectively represent the first, second and the th eigen-component coding vectors of the displacement fusion feature of the unit axis measurement.

[0052] That is, the pixel-by-pixel granular information association matrix is ​​decomposed into an orthogonal basis through characteristic spectrum decomposition. Its essence is to extract the inherent pattern of the pixel-by-pixel granular information association matrix, that is, by solving the eigenvalues ​​and eigenvectors of the pixel-by-pixel granular information association matrix, the original high-dimensional correlation data is projected into a low-dimensional linear space composed of principal components. Specifically, through this mathematical reconstruction method, the unstructured pixel-by-pixel granular information association matrix is ​​transformed into a set of atomic patterns with physical interpretability, providing a more discriminative structured feature basis for subsequent state judgment. The generated set of unit axis measurement displacement fusion feature eigencomponent encoding vectors, on the one hand, suppresses redundant noise by retaining high-energy principal components and enhances the model's sensitivity to systematic deviations of the axis. On the other hand, the linear independence of the orthogonal basis eliminates the collinear interference between features, so that the subsequent modules can accurately identify the evolution stage of the axis tortuosity based on the decoupled atomic pattern combination.

[0053] In the embodiment of the present application, each unit axis measurement displacement fusion feature intrinsic component coding vector in the set of the unit axis measurement displacement fusion feature intrinsic component coding vector is input into a feature significant modulation unit based on the self-attention mechanism to obtain a set of unit axis measurement displacement fusion feature intrinsic component significant modulation coding vectors, which can be expressed as follows:

[0054] in, represents a sequence model based on the self-attention mechanism, represents the set of significant modulation coding vectors of the fusion characteristic eigencomponents of the unit axis displacement measurement, , , Represents the first, second and The displacement fusion characteristic eigencomponents of the axis measurements of each unit significantly modulate the coding vector.

[0055] That is, through the feature significant modulation unit based on the self-attention mechanism, the interdependence weights between all the unit axis measurement displacement fusion feature intrinsic components in the set of unit axis measurement displacement fusion feature intrinsic component coding vectors are calculated to adaptively enhance the representation strength of the structural mode sensitive to the current fault, while weakening the interference of irrelevant or redundant components. The generated set of unit axis measurement displacement fusion feature intrinsic component significant modulation coding vectors significantly improves the model's perception sensitivity to weak but critical fault signals, and enhances the system's robustness under complex working conditions by dynamically suppressing the low-significance components dominated by noise, avoiding the accumulation of misjudgments caused by fixed feature weights.

[0056] In the embodiment of the present application, each unit axis measurement displacement fusion feature eigencomponent significant modulation coding vector in the set of the unit axis measurement displacement fusion feature eigencomponent significant modulation coding vector is projected onto the unit axis measurement displacement fusion feature kernel domain nonlinear activation matrix to obtain a set of unit axis measurement displacement fusion feature eigencomponent kernel mask coding vectors, which can be expressed as follows:

[0057] in, represents matrix multiplication, It represents the characteristic scale of the nonlinear activation matrix of the characteristic kernel domain of the unit axis measurement displacement fusion, Indicates The displacement fusion characteristic eigencomponent significant modulation coding vector of the axis measurement of each unit, It represents the length of the significant modulation coding vector of the eigencomponent of the displacement fusion characteristic of the unit axis measurement, Indicates The kernel mask encoding vector of the characteristic eigencomponent of the displacement fusion feature of each unit axis measurement is obtained.

[0058] That is, by projecting the significant modulation coding vector of the unit axis displacement fusion feature intrinsic component to the nonlinear activation matrix of the feature kernel domain, that is, by realizing the deep fusion of two heterogeneous information through generalized interactive operations, the system can perceive the data-driven hidden abnormal topology. In this way, a set of kernel mask coding vectors of the unit axis displacement fusion feature intrinsic component is generated, so that the model can still maintain the ability to accurately identify the axis tortuosity state and track stage evolution under complex working conditions.

[0059] In the embodiment of the present application, the set of kernel mask coding vectors of the unit axis measurement displacement fusion feature eigencomponents is fused to obtain the optimized unit axis measurement displacement fusion feature vector, which can be expressed as follows:

[0060] in, represents a cascade function, , , Represents the first, second and The kernel mask encoding vector of the displacement fusion feature eigencomponent of the axis measurement of each unit, Represents the fusion feature vector of optimized unit axis measurement displacement.

[0061] That is, by fusing the kernel mask encoding vectors of the eigencomponents of the axis measurement displacement fusion features of each unit, the limitations of single feature representation can be broken through. Through the selection of fusion strategy, the model can dynamically balance the diversity and simplicity of features according to actual needs, thereby generating an optimized unit axis measurement displacement fusion feature vector that contains multi-scale structural information and integrates complex context dependencies.

[0062] In an embodiment of the present application, the unit axis tortuosity state result generation module 140 is used to determine the tortuosity state of the hydropower unit axis based on the optimized unit axis measurement displacement fusion feature vector. Specifically, in an embodiment of the present application, the unit axis tortuosity state result generation module is used to: pass the optimized unit axis measurement displacement fusion feature vector through the unit axis tortuosity state classifier to obtain a tortuosity state classification result, and the tortuosity state classification result is used to indicate whether the hydropower unit axis is tortuous. It should be understood that the optimized unit axis measurement displacement fusion feature vector contains multi-scale and emphasized displacement features, which can fully and accurately reflect the state of the unit axis. However, these feature data are original and unprocessed, and cannot be directly used to determine whether the unit axis is tortuous. Therefore, it is necessary to further process and identify these features through the unit axis tortuosity state classifier to obtain accurate tortuosity state classification results. The unit axis tortuosity state classifier is a machine learning model that can analyze and judge based on the input data and map it to different categories. The results of the zigzag status classification directly reflect the operating status of the hydropower unit, helping engineers determine whether zigzags exist, so that maintenance measures can be taken in time to prevent the expansion of faults.

[0063] In summary, the hydroelectric generator unit axis measurement system 100 based on the embodiment of the present application is explained, which adopts artificial intelligence-based data analysis technology, and analyzes the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at multiple angles during the automatic turning process of the hydroelectric generator unit to determine whether the axis of the hydroelectric generator unit is tortuous. In this way, the displacement data at multiple angles can more comprehensively reflect the state of the axis of the hydroelectric generator unit, which is conducive to improving the accuracy and reliability of tortuosity judgment.

[0064] As described above, the hydroelectric unit axis measurement system 100 according to the embodiment of the present application can be implemented in various terminal devices, such as a server for measuring the axis of a hydroelectric unit. In one example, the hydroelectric unit axis measurement system 100 according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the hydroelectric unit axis measurement system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the hydroelectric unit axis measurement system 100 can also be one of the many hardware modules of the terminal device.

[0065] Alternatively, in another example, the hydropower unit axis measurement system 100 and the terminal device may also be separate devices, and the hydropower unit axis measurement system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

Claims

1. A hydroelectric unit axis measurement system, characterized in that: include: The hydropower unit related data acquisition module is used to obtain the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured by the laser sensor at multiple angles during the automatic turning of the hydropower unit; A hydropower unit related data sorting module is used to structure and regularize the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at the multiple angles to obtain a unit axis measurement displacement input matrix; A hydropower unit related data processing module is used to encode the unit axis measurement displacement input matrix to obtain an optimized unit axis measurement displacement fusion feature vector; The unit axis tortuosity state result generation module is used to judge the tortuosity state of the hydropower unit axis based on the optimized unit axis measurement displacement fusion feature vector.

2. The hydroelectric generator unit axis measurement system according to claim 1, characterized in that: The hydropower unit related data sorting module is used to: arrange the upper guide bearing X-direction and Y-direction displacement values, the lower guide bearing X-direction and Y-direction displacement values, the flange X-direction and Y-direction displacement values, and the water guide X-direction and Y-direction displacement values ​​measured at the multiple angles according to the angle dimension to obtain the unit axis measurement displacement input matrix.

3. The hydroelectric generator unit axis measurement system according to claim 2, characterized in that: The hydropower unit related data processing module includes: A unit axis measurement displacement feature extraction unit, used for performing preliminary feature extraction on the unit axis measurement displacement input matrix to obtain a multi-scale unit axis measurement displacement feature matrix; A unit axis measurement displacement feature emphasis unit, used for performing feature emphasis on the multi-scale unit axis measurement displacement feature matrix to obtain a unit axis measurement displacement emphasis feature matrix; The unit axis measurement displacement fusion feature generation unit is used to fuse the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain the optimized unit axis measurement displacement fusion feature vector.

4. The hydroelectric generator unit axis measurement system according to claim 3, characterized in that: The unit axis measurement displacement feature extraction unit is used to: pass the unit axis measurement displacement input matrix through a multi-scale axis measurement displacement feature extractor to obtain the multi-scale unit axis measurement displacement feature matrix.

5. The hydroelectric generator unit axis measurement system according to claim 4, characterized in that: The unit axis measurement displacement feature emphasis unit is used to: pass the multi-scale unit axis measurement displacement feature matrix through the axis measurement displacement feature emphasizer to obtain the unit axis measurement displacement emphasis feature matrix.

6. The hydroelectric generator unit axis measurement system according to claim 5, characterized in that: The axis measurement displacement feature emphasizer is a convolutional neural network model using a spatial attention mechanism, and the multi-scale axis measurement displacement feature extractor is a convolutional neural network model including a first convolutional layer and a second convolutional layer, wherein the first convolutional layer uses a two-dimensional convolutional kernel of a first scale, and the second convolutional layer uses a two-dimensional convolutional kernel of a second scale, and the first scale is different from the second scale.

7. The hydroelectric generator unit axis measurement system according to claim 6, characterized in that: The unit axis measurement displacement fusion feature generation unit comprises: A unit axis measurement displacement feature matrix dimension reduction subunit, used for expanding the multi-scale unit axis measurement displacement feature matrix and the unit axis measurement displacement emphasis feature matrix to obtain a multi-scale unit axis measurement displacement feature vector and a unit axis measurement displacement emphasis feature vector; A unit axis measurement displacement fusion feature generation subunit, used for weighted fusion of the multi-scale unit axis measurement displacement feature vector and the unit axis measurement displacement emphasis feature vector to obtain a unit axis measurement displacement fusion feature vector; The unit axis measurement displacement fusion feature optimization subunit is used to perform feature kernel domain projection modulation based on intrinsic regression on the unit axis measurement displacement fusion feature vector to obtain an optimized unit axis measurement displacement fusion feature vector.

8. The hydroelectric generator unit axis measurement system according to claim 7, characterized in that: The unit axis measurement displacement fusion feature optimization subunit is used to: Constructing a pixel-by-pixel granularity information association matrix of the unit axis measurement displacement fusion feature vector; Based on the convolution layer, kernel feature extraction is performed on the pixel-by-pixel granularity information association matrix to obtain a nonlinear activation matrix of the unit axis measurement displacement fusion feature kernel domain; Performing characteristic spectrum decomposition on the pixel-by-pixel granularity information association matrix to obtain a set of unit axis measurement displacement fusion characteristic eigencomponent encoding vectors; Input each unit axis measurement displacement fusion feature intrinsic component encoding vector in the set of unit axis measurement displacement fusion feature intrinsic component encoding vectors into a feature significant modulation unit based on a self-attention mechanism to obtain a set of unit axis measurement displacement fusion feature intrinsic component significant modulation encoding vectors; Projecting each unit axis measurement displacement fusion feature eigencomponent significant modulation coding vector in the set of the unit axis measurement displacement fusion feature eigencomponent significant modulation coding vectors to the unit axis measurement displacement fusion feature kernel domain nonlinear activation matrix to obtain a set of unit axis measurement displacement fusion feature eigencomponent kernel mask coding vectors; The set of kernel mask coding vectors of the unit axis measurement displacement fusion feature eigencomponents is fused to obtain the optimized unit axis measurement displacement fusion feature vector.

9. The hydroelectric generator unit axis measurement system according to claim 8, characterized in that: The unit axis tortuosity state result generation module is used to: pass the optimized unit axis measurement displacement fusion feature vector through the unit axis tortuosity state classifier to obtain a tortuosity state classification result, and the tortuosity state classification result is used to indicate whether the hydropower unit axis has tortuosity.

Images