A method for predicting and suppressing the risk of flutter in rotating machinery blades

Abstract

This invention discloses a method for predicting and suppressing the flutter risk of rotating machinery blades, belonging to the field of steam turbine operation safety technology. The method includes: acquiring a real-time operating condition parameter set of the steam turbine unit; inputting the real-time operating condition parameter set into a dynamic prediction model to establish the correlation between operating conditions and flutter risk, obtaining flutter risk prediction information; generating control commands based on the flutter risk prediction information, which are used to control a dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area; executing the control commands and collecting real-time vibration signals of the rotating machinery blades; and monitoring and adaptively controlling the steam injection parameters of the dynamic suppression flow channel based on the real-time vibration signals. This invention, employing a dynamic prediction model and a dual-channel cooling steam bypass system combined with flutter monitoring, can achieve real-time prediction, active suppression, and closed-loop adaptive control of steam turbine blade flutter risk.

Description

Technical Field

[0001] This invention relates to the field of turbine operation safety technology, and in particular to a method for predicting and suppressing the risk of blade flutter in rotating machinery. Background Technology

[0002] Rotating machinery is a core power source in key industrial sectors such as energy, chemical, and aviation. Large steam turbines are the main equipment in thermal power plants. Steam turbine blades, especially the long blades in the last few stages of the low-pressure cylinder, are prone to aeroelastic instability, or blade flutter, under specific operating conditions, such as low load or no-load operation, due to fluid-structure interaction. Blade flutter generates severe periodic vibrations, which can easily lead to fatigue damage to the blade material and even cause serious accidents such as blade fracture, posing a serious threat to the safe and stable operation of the unit.

[0003] Currently, the main control and protection strategies for turbine blade flutter are relatively conservative. One common approach is to set operating restriction zones, which uses control system logic to force the unit to avoid entering known low-load or high-vacuum operating ranges that are prone to inducing flutter. Another approach is to use continuous preventative cooling steam injection, which involves indiscriminately introducing a certain flow rate of cooling steam into the low-pressure cylinder when the unit enters potentially risky operating conditions, attempting to passively improve the blade operating environment by altering the overall flow field temperature and pressure.

[0004] However, the aforementioned existing technologies have the following drawbacks. Setting operating restriction zones sacrifices the unit's operational flexibility and load regulation capabilities, especially given the increasing demand for deep peak shaving in the power grid, which reduces the unit's economic efficiency. Furthermore, continuous preventative steam injection, due to a lack of accurate real-time risk assessment and flutter location, often leads to the blind and excessive use of steam. This not only wastes a large amount of high-quality steam and reduces the unit's cycle thermal efficiency, but its suppression effect is also unreliable and unable to cope with complex and changing actual operating conditions. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method for predicting and suppressing the risk of flutter in rotating machinery blades. By employing a dynamic prediction model and a dual-channel cooling steam bypass system, combined with flutter monitoring, it is possible to achieve real-time prediction, active suppression, and closed-loop adaptive control of turbine blade flutter risk.

[0006] The above objectives can be achieved through the following approach: A method for predicting and suppressing flutter risk in rotating machinery blades includes: acquiring a real-time operating parameter set of a steam turbine unit; inputting the real-time operating parameter set into a dynamic prediction model to establish the correlation between operating conditions and flutter risk, thereby obtaining flutter risk prediction information; generating control commands based on the flutter risk prediction information, wherein the control commands are used to control a dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area marker; executing the control commands and acquiring real-time vibration signals of the rotating machinery blades; and monitoring and adaptively controlling the steam injection parameters of the dynamic suppression flow channel based on the real-time vibration signals.

[0007] Optionally, acquiring the real-time operating parameter set of the turbine unit includes: collecting the inlet steam pressure, inlet steam temperature, exhaust steam vacuum, and rotor speed of the turbine unit to form macroscopic operating parameters; retrieving the blade aerodynamic load distribution map from a preset database storing multiple sets of load data corresponding to different macroscopic operating parameters based on the macroscopic operating parameters; and combining the macroscopic operating parameters and the blade aerodynamic load distribution map to generate a real-time operating parameter set.

[0008] Optionally, the step of inputting the real-time operating condition parameter set into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk to obtain flutter risk prediction information includes: matching and analyzing the real-time operating condition parameter set with historical operating condition data in the dynamic prediction model to calculate a flutter occurrence probability value that quantifies the current flutter risk; identifying the dominant vibration mode and circumferential location of historical flutter events that have a high correlation with the real-time operating condition parameter set; and integrating the flutter occurrence probability value, the dominant vibration mode, and the circumferential location to generate flutter risk prediction information that includes the predicted flutter risk level and the target sensitive area identifier.

[0009] Optionally, the step of generating control instructions based on the flutter risk prediction information includes: determining whether the predicted flutter risk level exceeds a preset risk threshold for triggering active intervention; if it does not exceed the threshold, generating a first control instruction to control the steady-state flow channel in the dual-channel cooling steam bypass system to open and maintain a basic steam flow rate; if it exceeds the threshold, generating a second control instruction to control the steady-state flow channel to open and simultaneously activate the dynamic suppression flow channel, wherein the second control instruction includes initial steam flow rate and injection angle parameters set for the activated dynamic suppression flow channel.

[0010] Optionally, the injection angle parameter includes: parsing the flutter risk prediction information and identifying the type of the dominant vibration mode; if the type of the dominant vibration mode is a bending mode, then the injection angle parameter is set to the steam jet direction that can generate radial damping force; if the type of the dominant vibration mode is a torsional mode, then the injection angle parameter is set to the steam jet direction that can generate reverse aerodynamic torque.

[0011] Optionally, the step of monitoring and adaptively controlling the steam injection parameters of the dynamic suppression channel based on the real-time vibration signal includes: extracting characteristic vibration amplitudes corresponding to the dominant vibration mode indicated in the flutter risk prediction information from the real-time vibration signal; calculating the deviation between the characteristic vibration amplitudes and a preset target amplitude range characterizing the safe vibration level of the blade; and generating a steam flow correction command for adjusting the dynamic suppression channel based on the deviation.

[0012] Optionally, the method further includes: after performing closed-loop fine-tuning and stabilizing the blade vibration, acquiring the final vibration response data of the blade; associating the real-time operating condition parameter set that triggered this protection, the flutter risk prediction information, and the final vibration response data to form a training data sample; and using the training data sample to update the internal parameters of the dynamic prediction model used to establish the correlation between operating conditions and flutter risk.

[0013] Optionally, generating the second control command includes: selecting a target nozzle covering the target sensitive area from the adjustable nozzle array based on the target sensitive area identifier; allocating the initial steam flow rate to the target nozzle based on the predicted flutter risk level; and setting the injection angle parameter in the target nozzle based on the dominant vibration mode.

[0014] Optionally, generating a steam flow correction instruction for adjusting the dynamic suppression channel based on the deviation includes: comparing the characteristic vibration amplitude with an upper threshold and a lower threshold of the target amplitude range; if the characteristic vibration amplitude is greater than the upper threshold, generating a first correction instruction for increasing the steam flow of the dynamic suppression channel, the first correction instruction including a flow increase calculated based on the deviation magnitude; if the characteristic vibration amplitude is less than the lower threshold, generating a second correction instruction for decreasing the steam flow of the dynamic suppression channel, the second correction instruction including a flow decrease calculated based on the deviation magnitude; and if the characteristic vibration amplitude is within the target amplitude range, generating a third correction instruction for maintaining the current steam flow of the dynamic suppression channel.

[0015] Based on the same inventive concept, this invention also provides a system for predicting and suppressing the flutter risk of rotating machinery blades. The system includes: a condition parameter acquisition module for acquiring real-time condition parameter sets of a steam turbine unit; a flutter risk prediction module for inputting the real-time condition parameter sets into a preset dynamic prediction model for establishing a correlation between condition and flutter risk, obtaining flutter risk prediction information, including a predicted flutter risk level and a target sensitive area identifier; a control command generation module for generating control commands based on the flutter risk prediction information, the control commands controlling a dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area identifier, reconstructing the flow field distribution within the low-pressure cylinder; a control command execution module for executing the control commands and acquiring real-time vibration signals of the rotating machinery blades; and a vibration monitoring and closed-loop control module for monitoring and adaptively controlling the steam injection parameters of the dynamic suppression flow channel based on the real-time vibration signals.

[0016] Compared with the prior art, the present invention has the following advantages: This invention achieves proactive and forward-looking management of flutter risk by constructing a closed-loop control system of prediction, intervention, and feedback. Compared with traditional passive response or fixed-condition avoidance strategies, this method can identify risks and initiate interventions before flutter occurs or worsens, thereby elevating safety margins from post-event remediation to pre-event prevention, enhancing the operational reliability and safety of the unit.

[0017] This invention achieves precision and efficiency in suppression measures. By introducing a dual-channel steam bypass system and combining it with the analysis of target sensitive area markers, the system can perform targeted steam injection into specific flow field regions. Simultaneously, the precise setting of the injection angle based on the dominant vibration mode ensures that the intervention energy is converted into aerodynamic damping, avoiding the significant steam waste and thermal efficiency loss associated with traditional large-scale, indiscriminate cooling methods, thus improving the economy and effectiveness of control.

[0018] This invention achieves dynamic optimization of suppression intensity by introducing closed-loop feedback and adaptive control based on real-time vibration signals. The system can fine-tune the steam injection parameters in real time according to the actual vibration level, ensuring that the suppression force is just right. This effectively suppresses blade vibration within a safe range while avoiding resource waste caused by excessive suppression, keeping the entire suppression process in a state of dynamic equilibrium optimization.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for predicting and suppressing the risk of flutter in rotating machinery blades according to an embodiment of the present invention.

[0022] Figure 2 This is a spatial mapping diagram of real-time operating conditions and load intensity in an embodiment of the present invention.

[0023] Figure 3 This is a nozzle selection diagram based on local flow field reconstruction in a flutter-sensitive region according to an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the structure of a system for predicting and suppressing the risk of flutter on rotating machinery blades, according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes a method for predicting and suppressing the flutter risk of rotating machinery blades. By employing a dynamic prediction model and a dual-channel cooling steam bypass system, combined with flutter monitoring, it is possible to achieve real-time prediction, active suppression, and closed-loop adaptive control of turbine blade flutter risk.

[0027] The method described in this embodiment specifically includes: S1. Obtain the real-time operating parameter set of the steam turbine unit; Optionally, acquiring the real-time operating parameter set of the steam turbine unit includes: Collect the steam inlet pressure, steam inlet temperature, exhaust vacuum, and rotor speed of the steam turbine unit to form macroscopic operating parameters; Based on the macroscopic operating parameters, the blade aerodynamic load distribution map is retrieved from a pre-set database that stores multiple sets of load data corresponding to different macroscopic operating parameters. The macroscopic operating parameters and the blade aerodynamic load distribution map are combined to generate a real-time operating condition parameter set.

[0028] Specifically, through a data acquisition interface, four key macroscopic operating parameters are collected in real time from the distributed control system (DCS) or monitoring information system (SIS) of the turbine unit at a frequency of 1 Hz to 10 Hz. These parameters include the inlet steam pressure and temperature, which directly affect the thermodynamic properties and energy input of the steam; the exhaust vacuum, which determines the work capacity; and the rotor speed, which characterizes the mechanical dynamics of the unit. These four parameters together constitute a macroscopic operating parameter vector, providing a basic basis for initially determining the unit's operating range.

[0029] Based on macroscopic operating conditions, the microscopic aerodynamic load distribution on the surface of rotating machinery blades is accurately inferred. Since directly deploying sensor arrays on high-speed rotating blades to measure aerodynamic loads is difficult, this method employs an indirect acquisition approach based on database queries. A blade aerodynamic load distribution map database is pre-established, generated through thousands of computational fluid dynamics (CFD) simulations or high-precision wind tunnel tests under different operating conditions. Each record in the database uniquely maps a specific set of macroscopic operating parameters to a detailed blade aerodynamic load distribution map. The mapping space between real-time operating conditions and load intensity is as follows: Figure 2 As shown, the blade aerodynamic load distribution map is a structured dataset that describes the pressure and aerodynamic force distribution experienced by the blade at different blade heights and on the blade tip and underside in a gridded form. It is crucial for revealing the aerodynamic excitation sources of blade flutter. Once real-time macroscopic operating parameters are obtained, they are used as an index to search and match within the database. Multidimensional interpolation algorithms may be employed to obtain the blade aerodynamic load distribution map that best matches the current operating conditions.

[0030] The data acquired in the first two steps are then structurally combined. This step is not a simple numerical concatenation, but rather the formation of a unified data structure. This structure includes both macroscopic operating parameters reflecting the overall energy and dynamic level of the unit, and blade aerodynamic load distribution maps describing the local stress state of the blades. This combination process can be represented by the following digital formula: , in, This represents the final set of real-time operating parameters generated. This refers to the steam inlet pressure of the unit. This refers to the steam inlet temperature of the unit. This refers to the exhaust vacuum of the generator unit. This refers to the rotor speed. These four items are macroscopic operating parameters that were collected in real time during the aforementioned steps. The aerodynamic load distribution map of the blade retrieved from the database is essentially a high-dimensional matrix or tensor, whose internal elements are the pressure or aerodynamic force values ​​at each node on the blade surface. The real-time operating parameter set generated in this way provides complete and profound physical feature inputs for subsequent dynamic prediction models to accurately assess blade flutter risk.

[0031] For example, the system first acquires four macroscopic operating parameters of the steam turbine unit in real time at a frequency of 10 Hz through a data acquisition interface. The currently measured inlet steam pressure is set to 0.8 MPa, inlet steam temperature to 180 degrees Celsius, exhaust steam vacuum to -92 kPa, and rotor speed to 3000 rpm. The system then uses these four macroscopic operating parameters as indexes to search a pre-set aerodynamic load distribution map database containing thousands of computational fluid dynamics simulation results. During database matching, the system locates the load dataset closest to the current macroscopic operating parameters and extracts the corresponding blade surface pressure distribution tensor. This tensor details the stress state of each grid node from the blade root to the blade tip of the last-stage blade. Finally, the system integrates the above parameters through structured combination. Calculations are performed according to digital formulas, substituting the measured values ​​to obtain... ,in This is a high-dimensional pressure matrix retrieved from the database. In this way, the system successfully generated a complete set of real-time operating parameters, providing core physical feature inputs for subsequent dynamic prediction models to accurately assess blade flutter risk.

[0032] S2. Input the real-time operating condition parameter set into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk to obtain flutter risk prediction information, which includes the predicted flutter risk level and the target sensitive area identifier. Optionally, the step of inputting the real-time operating condition parameter set into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk to obtain flutter risk prediction information includes: The real-time operating condition parameter set is matched and analyzed with the historical operating condition data in the dynamic prediction model to calculate the flutter occurrence probability value that quantifies the current flutter risk. Identify the dominant vibration mode and circumferential location of historical flutter events that are highly correlated with the real-time operating condition parameter set; By integrating the flutter occurrence probability value, the dominant vibration mode, and the circumferential location of occurrence, flutter risk prediction information is generated, which includes the predicted flutter risk level and the target sensitive area identifier.

[0033] Specifically, within a multidimensional feature space composed of historical data, the current operating condition is located and its proximity to known flutter events is assessed. The real-time operating condition parameter set generated by the preceding steps is received and input as a query vector into the dynamic prediction model. The dynamic prediction model is a hybrid machine learning model that integrates an improved K-nearest neighbor algorithm with a lightweight fully connected neural network. It features a lightweight four-layer architecture: feature extraction, operating condition matching, risk mapping, and modal recognition. Single-frame processing time is ≤100ms, making it suitable for deployment on conventional industrial control computers. The feature extraction layer reduces the dimensionality of the aerodynamic load map and concatenates it with 4-dimensional macroscopic operating parameters to form a 132-dimensional fused feature vector. After standardization, this vector is used by the operating condition matching layer to calculate similarity using a K-nearest neighbor algorithm with 20 neighbors. The risk mapping layer completes flutter probability mapping through a three-layer fully connected neural network and then classifies the risk into low, medium, and high risk levels based on probability values. The modal recognition layer determines the dominant vibration mode based on K-nearest neighbor voting and determines the circumferential position by combining the statistical analysis of neighboring operating condition positions, achieving a positioning accuracy of ≤10 degrees. The model weight matrix is ​​calibrated offline using grey relational analysis. Parameter updates employ stochastic gradient descent with a mean squared error loss function and adaptive learning rate, ceasing updates once a specific convergence condition is met. Model training relies on a database of over 10,000 CFD simulations and real-machine tests. After three levels of preprocessing, the data is divided into training, validation, and test sets in an 8:1:1 ratio. After offline training, the model must achieve predetermined prediction and recognition accuracy before being exported in a lightweight format for deployment. The industrial-grade model is adapted to the turbine generator DCS / SIS system, interacting via the OPC UA protocol. A dual-machine redundancy architecture ensures stable operation, with strict limits on data transmission and output latency. The host computer monitors and issues alarms for any anomalies. The model is maintained and updated daily according to established rules. Training samples are generated upon completion of single-cycle flutter suppression, and parameters are automatically updated upon reaching a quantitative threshold. Offline retraining is performed annually, and accuracy is verified quarterly; if standards are not met, immediate retraining is initiated. The model connects with other modules of the system via TCP / IP protocol and JSON format, receiving operating condition parameters as input and outputting flutter risk prediction information. It also receives vibration response feedback data as training samples. The model internally stores a large amount of historical operating condition data; each data point contains a historical operating condition parameter set and an associated label, which explicitly indicates whether flutter occurred under that operating condition. A weighted distance algorithm is used to calculate the similarity between the current real-time operating condition parameter set and each historical operating condition data point in the model database. This similarity... The calculation method is as follows: , In this formula, It represents the feature vector derived from the current real-time operating condition parameter set, including macroscopic operating parameters and blade aerodynamic load distribution map. This represents the feature vector corresponding to a specific historical operating condition data point. This is a diagonal weight matrix, where the elements on the diagonal reflect the sensitivity of different operating parameters to flutter. These weight coefficients are determined through offline training and expert experience. By calculating the minimum distance between the current operating condition and all historical flutter event conditions, a basis for quantifying the current flutter risk is obtained. Subsequently, this minimum distance is converted into a flutter probability value between 0 and 1 using a preset probability mapping function, typically the sigmoid function. This probability value intuitively reflects the likelihood of blade flutter under the current operating condition.

[0034] Identify the most likely vibration mode and spatial location of flutter. After similarity calculation, select one or more sets of historical flutter events with the highest similarity to the current operating condition. These historical flutter event data records, in addition to operating parameters, also store detailed flutter characteristics captured by the vibration monitoring system at the time, namely the dominant vibration mode and the circumferential location of occurrence. The dominant vibration mode refers to the specific vibration pattern in which the blade vibration energy is most concentrated during flutter, such as the first bending mode or the first torsional mode, which determines the main deformation mode of the blade. The circumferential location of occurrence indicates the specific sector of the flutter phenomenon within the 360-degree circumferential range of the low-pressure cylinder, for example, within the range of 0 to 45 degrees. Statistical analysis of the characteristics of the selected high-similarity historical flutter events, such as using a voting mechanism or weighted averaging, is performed to determine the most likely dominant vibration mode and circumferential location under the current risk.

[0035] All analysis results are integrated to form a structured and actionable flutter risk prediction information. The flutter occurrence probability values, predicted dominant vibration modes, and circumferential locations obtained in the first two steps are integrated. First, based on the flutter occurrence probability values ​​and preset risk level thresholds (e.g., flutter occurrence probability values ​​less than 0.3 indicate low risk, 0.3 to 0.7 indicate medium risk, and greater than 0.7 indicate high risk), they are mapped to specific predicted flutter risk levels. Then, the identified dominant vibration modes and circumferential locations are combined to form target sensitive area identifiers. The final generated flutter risk prediction information includes both the severity of the risk and its specific physical manifestation and spatial location, providing precise target guidance for the generation of subsequent control commands.

[0036] For example, after acquiring the real-time operating condition parameter set, the system inputs it as a query vector into the dynamic prediction model to quantify the current flutter risk. Assume the currently extracted feature vector... This is a set of normalized values ​​containing macroscopic operating parameters and blade aerodynamic load characteristics, and a typical historical flutter condition feature vector stored in the model database. Given, and pre-defined, diagonal weight matrix This is used to demonstrate the sensitivity of different parameters to the impact of flutter. The system employs a weighted distance algorithm to calculate the similarity between the current operating condition and historical flutter events. ,set up , , Calculate vector difference Substituting into the formula yields the similarity distance. The system then uses a mapping function to convert this distance into a flutter probability value. Assuming the mapping result is 0.82, which exceeds a preset high-risk threshold, the system maps it to a high-risk level. Next, the system filters out this historical operating condition. The associated flutter features identified the dominant vibration mode as a first-order bending mode, with circumferential locations within the 45- to 90-degree sector. Finally, the system integrates the flutter occurrence probability value, dominant vibration mode, and circumferential location to generate flutter risk prediction information that includes the predicted flutter risk level and target sensitive area identification, providing target guidance for subsequent precise steam injection.

[0037] S3. Generate control instructions based on the flutter risk prediction information. The control instructions are used to control the dual-channel cooling steam bypass system with steady-state flow channel and dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area identifier, and reconstruct the flow field distribution in the low-pressure cylinder. Optionally, generating control instructions based on the flutter risk prediction information includes: Determine whether the predicted flutter risk level exceeds a preset risk threshold for triggering active intervention; If the limit is not exceeded, a first control command is generated to control the opening of the steady-state flow channel in the dual-channel cooling steam bypass system and maintain the basic steam flow rate. If the limit is exceeded, a second control command is generated to control the opening of the steady-state flow channel and simultaneously activate the dynamic suppression flow channel. The second control command includes the initial steam flow rate and injection angle parameters set for the activated dynamic suppression flow channel.

[0038] Specifically, a clear and automated intervention initiation decision-making mechanism is established. The control system first analyzes the flutter risk prediction information output by the dynamic prediction model, extracting the core predicted flutter risk level, which is typically represented as a flutter occurrence probability value between 0 and 1. The system compares this probability value with a pre-set risk threshold used to trigger active intervention. This risk threshold is a key safety parameter determined based on extensive experimental data, simulation analysis, and operational experience; its value is usually set between 0.6 and 0.8, representing the upper limit of risk that the system can tolerate. This comparison operation is the trigger point for the entire active protection logic, determining whether the system switches from the conventional cooling mode to the active suppression mode.

[0039] When the flutter risk is low, basic preventative cooling measures are implemented. If the predicted flutter risk level does not exceed the risk threshold, it indicates that although there is a certain tendency for flutter in the current operating condition, it is still within a controllable range and no strong intervention is required. At this time, the first control command will be generated. The goal of this command is to control the opening of the steady-state flow channel in the dual-channel cooling steam bypass system and adjust its valve opening according to the current macroscopic operating parameters, such as exhaust steam temperature, to maintain a basic steam flow rate. The steady-state flow channel is designed with uniform nozzle distribution to form a stable, low-speed steam curtain surrounding the blade flow channel. Its main function is to cool the last-stage blades and improve the flow uniformity at the blade inlet, thereby optimizing the flow field environment and suppressing the initiation of flutter without consuming bypass steam.

[0040] When the flutter risk exceeds the safety threshold, powerful and precise active suppression measures are immediately initiated. If the predicted flutter risk level exceeds the risk threshold, the system determines that blade flutter is about to occur or is already in its nascent stage, requiring active intervention. The system immediately generates a second control command. This command contains more complex control logic. It first ensures that the steady-state flow channel remains open to continue providing basic cooling, while the core action is to activate the dynamic suppression flow channel. The dynamic suppression flow channel consists of a series of high-speed-response electronically controlled valves and adjustable nozzles, enabling instantaneous, targeted, and directional injection of high-pressure steam. The second control command includes two key initial parameters set for the activated dynamic suppression flow channel: the initial steam flow rate and the injection angle parameter. The initial steam flow rate is not a fixed value but is positively correlated with the overshoot of the risk, and its calculation method can be expressed as: , in, This represents the initial steam flow rate set. It is the minimum activation flow rate required for the dynamic suppression channel to effectively generate aerodynamic damping. It is a gain coefficient, obtained through system calibration. It is the predicted flutter risk level. This is the risk threshold. This formula ensures that the higher the risk, the greater the initial intervention. Meanwhile, the injection angle parameter included in the instruction is precisely set based on the target sensitive area identification and dominant vibration mode in the flutter risk prediction information, aiming to most effectively apply the energy of the steam jet to suppress specific modes of flutter.

[0041] For example, after receiving flutter risk prediction information, the control system first extracts the predicted flutter risk level and compares it with a preset risk threshold for automated decision-making. A preset risk threshold is set to trigger active intervention. The value is 0.70, while the flutter occurrence probability value output by the current dynamic prediction model is the predicted flutter risk level. It is 0.85. Because... Greater than The system determines that the risk has exceeded the safety threshold and immediately generates a second control command to activate the dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel. During the command generation process, the system needs to calculate the initial steam flow rate of the dynamic suppression flow channel. Set the minimum activation flow rate required to dynamically suppress aerodynamic damping in the flow channel. The gain is 5.0 kg / s. The value was calibrated and set to 20. The result was obtained by substituting the example data into the digitization formula. Calculations showed that the initial steam flow rate was set to 8.0 kg / s. Simultaneously, the steady-state flow channel was kept open, and the base steam flow rate was maintained for preventative cooling. Furthermore, the injection angle parameters were synchronously set based on the target sensitive area identification and the dominant vibration mode. Finally, a complete second control command containing the initial steam flow rate and injection angle parameters was sent to the actuator.

[0042] Optionally, the injection angle parameters include: Analyze the flutter risk prediction information to identify the type of the dominant vibration mode; If the dominant vibration mode is a bending mode, then the injection angle parameter is set to the direction of the steam jet that can generate radial damping force; If the dominant vibration mode is a torsional mode, then the injection angle parameter is set to the direction of the steam jet that can generate reverse aerodynamic torque.

[0043] Specifically, the key features determining the blade vibration pattern are precisely extracted from the flutter risk prediction information. When generating the second control command, the control system first performs a deep analysis of the flutter risk prediction information generated in the previous steps. This information includes the identification results of the most likely dominant vibration mode under the current operating condition. The dominant vibration mode is a manifestation of the blade's structural dynamic characteristics, and different modes correspond to different deformation patterns of the blade. For example, the bending mode is mainly characterized by the flapping motion of the blade along the blade height direction, while the torsional mode is characterized by the torsional vibration of the blade around a certain axis. By reading the mode type identifier, the system provides input for subsequent differentiated injection strategy selection.

[0044] When the predicted dominant vibration mode is the bending mode, the injection angle that provides the maximum radial damping force is set. Bending mode flutter manifests as the blade's reciprocating motion along the radial or axial direction. To suppress this motion, a force in the opposite direction or capable of effectively dissipating its kinetic energy is required. Accordingly, the system sets the injection angle parameter of the target nozzle in the dynamic suppression channel to a direction that generates radial damping force. Radial damping force is an aerodynamic force perpendicular to the blade's flapping motion direction, directly suppressing its vibration amplitude. This means that the steam jet direction typically forms a large angle with the tangent of the blade's arc at the injection point, such as 45 to 90 degrees, allowing the impact and entrainment effects of the steam jet to maximize their effect on hindering the blade's bending deformation, thereby rapidly attenuating its vibrational energy.

[0045] When the predicted dominant vibration mode is torsional, an injection angle is set to generate a suppressive aerodynamic torque. Torsional flutter manifests as torsional oscillations of the blade around its aerodynamic center, driven by an unstable coupling between changes in the blade angle of attack and aerodynamic forces. The most effective way to suppress this torsional vibration is to apply an aerodynamic torque opposite to the direction of the flutter driving torque. Therefore, the system sets the injection angle parameter to a steam jet direction that generates a reverse aerodynamic torque. Specifically, this typically requires the steam jet direction to be close to the blade chord direction, and the injection position to be off-center from the blade pressure center, such as at the leading or trailing edge. In this way, the high-speed steam jet creates an asymmetric pressure distribution on the blade surface, generating a net aerodynamic torque. The direction of this torque, after precise calculation, can counteract the unstable aerodynamic torque that causes torsional flutter, thereby disrupting the self-excited conditions for flutter and restoring the torsional stability of the blade.

[0046] For example, during the generation of the second control command, the control system first accurately identifies the type of dominant vibration mode by analyzing flutter risk prediction information. This step aims to customize the most effective aerodynamic suppression scheme based on the different physical deformation characteristics of the blade. If the system identifies the dominant vibration mode of the current blade as a bending mode, determining that the blade is in a reciprocating flapping motion along the radial or axial direction, the system will automatically set the injection angle parameter to the direction that can generate the maximum radial damping force. In physical implementation, the system controls the adjustable nozzle to make the steam jet direction form a large angle with the tangent direction of the blade's arc, using the impact and entrainment effect generated by the jet to hinder the bending deformation of the blade, thereby quickly dissipating its vibration energy. If the analysis result shows that the dominant vibration mode is a torsional mode, indicating that the blade is undergoing torsional oscillation around its aerodynamic center, the system will switch the injection angle parameter to the direction that can generate the reverse aerodynamic torque. In specific operation, the system adjusts the target nozzle to make the steam jet direction close to the blade chord direction and ensures that the injection position is deviated from the pressure center of the blade, for example, positioned in the leading or trailing edge region of the blade. This precise directional injection creates an asymmetric pressure distribution on the blade surface, which in turn generates a net aerodynamic torque sufficient to counteract the flutter driving torque. By disrupting the self-excited conditions that cause flutter, the blade regains torsional stability.

[0047] Optionally, generating the second control command includes: Based on the target sensitive area identifier, select the target nozzles that cover the area from the adjustable nozzle array; Based on the predicted flutter risk level, the initial steam flow rate is allocated to the target nozzle; The injection angle parameter is set for the target nozzle based on the dominant vibration mode.

[0048] Specifically, nozzle selection based on local flow field reconstruction in flutter-sensitive regions, such as... Figure 3 As shown, the abstract risk area location is transformed into the precise selection of physical actuators. The control system first parses the target sensitive area identifier from the flutter risk prediction information. This identifier is a data structure containing circumferential start and end angles, such as a sector from 30 degrees to 75 degrees, and possibly a radial blade height range. It then invokes a preset nozzle mapping database, which stores the precise three-dimensional spatial coordinates of each nozzle in the entire adjustable nozzle array contained within the dynamic suppression channel. By executing a spatial geometry matching algorithm, the spatial range defined by the target sensitive area identifier is compared with the effective coverage range of all nozzles, filtering out all nozzles capable of effectively projecting steam jets into that area, forming a list or set of target nozzles.

[0049] Based on the urgency of the risk, a reasonable initial intervention intensity is assigned to the selected target nozzle group. The system calculates the required total initial steam flow rate based on the predicted flutter risk level, i.e., the probability of flutter occurrence. This total flow rate is then distributed to all the target nozzles selected in the previous step. The most straightforward distribution strategy is average distribution, where each target nozzle receives the same initial flow rate setpoint. This distribution process can be described by the following formula: , in, Is assigned to the first Initial steam flow rate of each target nozzle. It is the total initial steam flow rate calculated based on the risk level. This refers to the number of target nozzles selected. This ensures that the overall intervention intensity matches the risk level and is applied evenly across the entire sensitive area.

[0050] To ensure the applied intervention force is optimal in direction and maximizes the suppression of specific flutter patterns, the final step after allocating flow to the target nozzles is to set their injection angle parameters. Based on the dominant vibration mode type identified in the flutter risk prediction information, uniform or differentiated injection angles are set for all target nozzles. The setting logic is as follows: for bending modes, an angle that generates maximum radial damping force is set; for torsional modes, an angle that generates reverse aerodynamic torque is set. Finally, the complete instruction set, including the target nozzle ID, its initial steam flow value, and injection angle parameters, is packaged into a second control command and sent to the controller of the dual-channel cooling steam bypass system, thereby achieving a rapid, precise, and targeted response to flutter risk.

[0051] For example, when generating the second control command, the control system first parses the target sensitive area identifier in the flutter risk prediction information. This identifier defines the circumferential sector where the current risk occurs, setting this area to a range of 30 to 75 degrees circumferentially. The system calls a mapping database storing the precise three-dimensional coordinates of the adjustable nozzle array and, by executing a spatial geometric matching algorithm, filters out target nozzles from all nozzles that can effectively cover this 30 to 75 degree sector. Assuming the number of selected target nozzles... There are four. The system then calculates the required total initial steam flow rate based on the predicted flutter risk level. The value was set to 12.0 kg / s. To ensure the intervention intensity was applied evenly to the sensitive area, the system performed an initial steam flow distribution calculation. The calculation was performed according to the formula... The calculation results show that the initial steam flow rate allocated to each target nozzle is... All are 3.0 kg / s. Finally, based on the dominant vibration mode of flutter, the system synchronously sets uniform injection angle parameters for these four target nozzles, thereby completing the encapsulation of the second control command and sending it to the dual-channel cooling steam bypass system.

[0052] S4. Execute the control command to collect the real-time vibration signal of the rotating mechanical blade; Specifically, after receiving and executing control commands, the system enters the feedback monitoring phase. The control system uses timing sensors pre-installed on the blade tips of the turbine casing or strain gauges attached to the blade roots to dynamically capture data from the high-speed rotating mechanical blades at a kilohertz sampling rate, thereby acquiring raw electrical signals reflecting the mechanical vibration state of the blades. To extract effective information from the complex signal containing noise and multimodal interference, the raw electrical signal needs to be pre-processed in real time. A fast Fourier transform is used to convert the time-series signal into frequency domain features, and a high-precision digital bandpass filter is applied to filter it based on the pre-identified dominant vibration mode frequency range from the flutter risk prediction information. The filtered signal is quantized into characteristic vibration amplitude, which directly characterizes the degree of blade deformation under the target mode. A preset target amplitude range, consisting of an upper and lower threshold, is then introduced to evaluate the effectiveness of the current flow field reconstruction measures. By calculating the deviation between the characteristic vibration amplitude and the target safety boundary, feedback input is provided for subsequent flow fine-tuning decisions. In practice, the characteristic vibration amplitude is obtained by envelope detection of the peak value of the filtered vibration signal, while the upper threshold is a safe operating boundary value determined based on the fatigue limit of the blade material and finite element strength simulation. Through this high-frequency real-time sampling and signal demodulation, the microscopic vibration physical state of the blade can be transformed into a digital feedback index, providing a scientific basis for the fine-tuning of dynamic suppression of steam parameters in the flow channel.

[0053] S5. Based on the real-time vibration signal, monitor and adaptively control the steam injection parameters of the dynamic suppression channel.

[0054] Optionally, the step of monitoring and adaptively controlling the steam injection parameters of the dynamic suppression channel based on the real-time vibration signal includes: Extract the characteristic vibration amplitude corresponding to the dominant vibration mode indicated in the flutter risk prediction information from the real-time vibration signal; Calculate the deviation between the characteristic vibration amplitude and the preset target amplitude range characterizing the safe vibration level of the blade; Based on the deviation, a steam flow correction command is generated to adjust the dynamic suppression channel.

[0055] Specifically, the vibration components directly related to the current suppression target are precisely separated from the complex real-time vibration signal. After the dynamic suppression channel is activated, real-time vibration signals of the rotating machinery blades are continuously acquired at a kilohertz sampling rate using a timed BTT sensor mounted on the blade tip on the casing or a strain gauge attached to the blade root. Since blade vibration is the result of multiple modes superimposed, the raw signal must be processed in real time. Using the dominant vibration mode indicated by the flutter risk prediction information in the previous steps, a high-precision digital bandpass filter or a Fast Fourier Transform (FFT) is applied to extract the energy corresponding to the specific mode frequency from the broadband vibration signal. This energy is quantized as a characteristic vibration amplitude, which directly reflects the intensity of the target flutter mode and is the sole feedback quantity for subsequent closed-loop control.

[0056] The system quantifies the gap between the current vibration suppression effect and the preset safety target. It compares the characteristic vibration amplitude calculated in real time with a preset target amplitude range characterizing the safe vibration level of the blade. This target amplitude range is not a single value, but a safe interval including an upper and lower limit; for example, the blade tip amplitude is between 50 and 80 micrometers. This range is determined through finite element analysis of blade dynamics and fatigue life assessment. The upper limit represents the dangerous vibration threshold that must be avoided, while the lower limit is the economical operating threshold set to avoid excessive consumption of bypass steam while ensuring safety. The system calculates the difference between the characteristic vibration amplitude and this range boundary to obtain a quantified deviation, which directly drives subsequent adjustment decisions.

[0057] Based on the magnitude and direction of the deviation, a correction command is generated for fine-tuning the steam injection rate. This process employs a proportional-integral PID control algorithm to ensure both speed and stability of the adjustment. The deviation calculated in the previous step is used as the input to the PID controller to generate a correction amount for the steam flow rate in the dynamically suppressed flow channel. The calculation of this correction amount can be simplified to the following proportional control logic: , in, It is the calculated steam flow correction command value. It is a tuned proportional gain coefficient that determines the response sensitivity of the control system. This is the deviation between the characteristic vibration amplitude and the boundary of the target amplitude range. If the characteristic vibration amplitude exceeds the upper limit of the target range, If the value is positive, the generated The command will increase the steam flow rate to enhance the suppression effect; if the characteristic vibration amplitude is below the lower limit of the target range, For negative values, the generated The instruction will reduce steam flow to save energy; if the amplitude is within the target range, If the value is zero, the current flow rate is maintained. This correction command is ultimately sent as a digital or analog signal to the flow control valve actuator in the dynamic suppression channel, thus completing one closed-loop fine-tuning. The entire closed-loop process is executed cyclically at a frequency of 1 to 5 Hz until the blade vibration stabilizes within the target amplitude range.

[0058] For example, after executing control commands, the control system continuously acquires real-time vibration signals of the rotating machinery blades at a sampling rate of kilohertz using timing sensors mounted on the blade tips of the casing or strain gauges attached to the blade roots. The system first utilizes the dominant vibration mode indicated by the flutter risk prediction information from the preceding steps, applying a digital bandpass filter or performing a fast Fourier transform to accurately separate specific modal components from the broadband vibration signal and quantize them into characteristic vibration amplitudes. The system then compares the currently measured characteristic vibration amplitude with a preset target amplitude range. If there is a deviation between the characteristic vibration amplitude and the target range boundary... The system then generates a steam flow correction command based on this deviation to adjust the dynamic suppression flow channel. The measured characteristic vibration amplitude is set at 95 micrometers, and the upper limit threshold of the safety target is set at 80 micrometers; the deviation between the two is... The value is 15 micrometers, and the proportional gain coefficient after system tuning is set. The value is 0.2. Based on the formula for calculating the correction amount, we obtain... The calculation results show that the system generates a correction command to increase the steam flow rate by 3.0 kg / s and sends it to the flow control valve actuator. The entire closed-loop process is executed cyclically at a frequency of 1 to 5 Hz until the blade vibration is stably controlled within the target amplitude range, thus ensuring the robustness and optimality of the protection measures.

[0059] Optionally, generating a steam flow correction command for adjusting the dynamic suppression channel based on the deviation includes: The characteristic vibration amplitude is compared with the upper and lower threshold values ​​of the target amplitude range; If the characteristic vibration amplitude is greater than the upper limit threshold, a first correction instruction is generated to increase the steam flow rate of the dynamic suppression channel. The first correction instruction includes the flow rate increase calculated based on the deviation magnitude. If the characteristic vibration amplitude is less than the lower threshold, a second correction instruction is generated to reduce the steam flow rate of the dynamic suppression channel. The second correction instruction includes the flow rate reduction calculated based on the deviation magnitude. If the characteristic vibration amplitude is within the target amplitude range, a third correction command is generated to maintain the current steam flow rate of the dynamic suppression channel.

[0060] Specifically, a clear, three-interval comparison logic is established as the basis for all adjustment actions. Within each cycle of the closed-loop control, the real-time acquired characteristic vibration amplitude is first compared with a preset target amplitude range. This target amplitude range is defined by two key parameters: an upper threshold and a lower threshold. These two thresholds together define a safe vibration "green zone." The comparison result has only three possibilities: the characteristic vibration amplitude is higher than the upper threshold, lower than the lower threshold, or falls between the two. This comparison step is the sole basis for generating different correction commands subsequently, ensuring that the decision-making logic of the control system is clear and unambiguous.

[0061] When vibration exceeds the limit, an incremental adjustment is executed to enhance the suppression effect. When the characteristic vibration amplitude is determined to be greater than the upper threshold, it indicates that the current steam injection rate is insufficient to effectively suppress flutter, and the blades remain in a dangerous state. At this point, a first correction command is immediately generated, the core of which is to increase the steam flow rate in the dynamic suppression channel. This command includes a specific flow rate increase, which is not a fixed value but is calculated based on the magnitude of the deviation using a proportional control law, ensuring the smoothness and appropriateness of the adjustment. The formula for calculating the flow rate increase is as follows: , in, It is the calculated increase in flow. It is a proportional gain coefficient specifically used for incremental adjustment. It is the characteristic vibration amplitude measured in real time. This is the upper limit threshold of the target amplitude range. This instruction is sent to the flow controller, causing it to open the valve wider and increase the suppression force.

[0062] When vibration is excessively suppressed, a reduction regulation aimed at saving energy is executed. If the characteristic vibration amplitude is determined to be below the lower threshold, it indicates that the current steam injection rate has exceeded the necessary level. Although the blades are perfectly safe, this results in unnecessary bypass steam consumption, affecting unit efficiency. Therefore, a second correction command is generated, the core of which is to reduce the steam flow in the dynamically suppressed flow channel. Similar to incremental regulation, the command includes a flow reduction calculated based on the deviation magnitude. , in, It is the calculated reduction in flow. It is the proportional gain coefficient used for deceleration adjustment, and its value may be related to... Different approaches are used to achieve asymmetric control. This is the lower threshold of the target amplitude range. This command will reduce the flow rate valve to seek a more economical operating point while ensuring safety.

[0063] When vibration is under ideal conditions, the system remains stable, avoiding unnecessary disturbances. If the characteristic vibration amplitude is neither too large nor too small, falling precisely within the target amplitude range (between the lower and upper thresholds), the system determines the current control state to be optimal. At this point, a third correction command is generated, essentially a "no-operation" or "hold" command, used to maintain the current steam flow rate of the dynamic suppression channel unchanged. This ensures system stability and avoids unnecessary oscillations within the target range due to overly sensitive control systems.

[0064] For example, during the closed-loop control cycle, the control system first extracts the characteristic vibration amplitude values ​​in real time from the sensors. The amplitude is compared with a preset target range, which is defined by an upper threshold. and lower threshold Common definition. Set the currently measured characteristic vibration amplitude. 95 micrometers, upper limit threshold 80 micrometers, lower threshold The value is 50 micrometers. Because the characteristic vibration amplitude exceeds the upper threshold, the system determines that the suppression force is insufficient and generates a first correction command to increase the flow rate. The proportional gain coefficient for incremental adjustment is set. The value is 0.2. This is calculated according to the formula. The calculation results show that the system generated a command to increase the steam flow rate by 3.0 kg / s. If the characteristic vibration amplitude under another operating condition... If the flow rate drops to 40 micrometers, below the lower threshold, the system determines this as excessive intervention and generates a second correction command to reduce the flow. The proportional gain coefficient for the reduction adjustment is set. It is 0.15. Calculated according to the formula... Calculations showed that the system generated a command to reduce the steam flow rate by 1.5 kg / s. If the characteristic vibration amplitude was between 50 and 80 micrometers, the system generated a third correction command to maintain the current flow rate. Through this asymmetric proportional adjustment based on the magnitude of the deviation, the system ensured that the blade vibration could be stabilized within a safe and economical operating range.

[0065] Optionally, the method further includes: After performing closed-loop fine-tuning and stabilizing the blade vibration, the final vibration response data of the blade is obtained; The real-time operating condition parameter set that triggered this protection, the flutter risk prediction information, and the final vibration response data are correlated to form a training data sample. The internal parameters of the dynamic prediction model used to establish the correlation between operating conditions and flutter risk are updated using the training data samples.

[0066] Specifically, after a successful flutter suppression event, the final system state related to that event is fully captured. Once the closed-loop fine-tuning process is complete—that is, after the blade vibration has been successfully stabilized within the target amplitude range for a period of time, such as exceeding 30 seconds—the system determines that the protection task is complete. At this point, a data acquisition program is triggered to collect the final vibration response data of the blade in this stable state. This data not only includes the average characteristic vibration amplitude after stabilization but may also include subtle shifts in vibration frequency, spectral distribution characteristics of the vibration signal, and other richer dynamic information. This final vibration response data provides direct evidence, from the results, verifying the effectiveness of the protection measures.

[0067] The input of the triggering event, the model's predicted output, and the final result of the intervention are structurally correlated to form a high-quality training sample. The entire chain of data from this protection event is then integrated. This process includes three key parts: first, the real-time operating parameter set at the moment the protection event was triggered, representing the "problem" leading to flutter risk; second, the flutter risk prediction information output by the dynamic prediction model at that time, representing the model's "diagnosis" of the "problem"; and finally, the newly collected final vibration response data, representing the "ideal result" achieved by the system after taking corresponding measures. These three parts of data are correlated and packaged to form a new, complete training data sample. This sample records a complete closed-loop process from risk identification to successful suppression, possessing extremely high learning value.

[0068] Leveraging newly acquired valuable experience, the internal parameters of the dynamic prediction model are updated online or offline to continuously improve its adaptability and prediction accuracy over time. Newly generated training data samples are fed into the model's training module. This module employs incremental learning or online learning algorithms in machine learning, such as stochastic gradient descent, to update the internal parameters of the dynamic prediction model used to establish the correlation between operating conditions and flutter risk. This update process aims to minimize the discrepancy between model predictions and actual results. For example, if the model predicts a high risk level, but vibration is ultimately suppressed with only a small steam flow rate, this indicates that the model may have overestimated the risk. The training algorithm will adjust the model parameters to reduce its sensitivity near that operating condition. This update process can be represented by the following general optimization formula: , In this formula, This represents the updated set of internal parameters of the model. This represents the parameters before the update. It is the learning rate, a hyperparameter that controls the step size for each update, and its value is usually between 0.001 and 0.1. These are newly generated training data samples. It is a loss function The gradient with respect to new samples under the old parameters indicates the direction of parameter adjustment so that the model can adapt to similar situations. When the operating conditions are favorable, the predicted results are closer to the actual system response. Through this continuous self-iteration, the dynamic prediction model can continuously learn new operating condition-risk-response relationships, thereby making its predictions more and more accurate, and the protection strategy is also optimized accordingly.

[0069] For example, after the blade vibration stabilizes, the system first extracts the data assets of the entire process of this protection event and integrates them into a set of complete training data samples. This sample includes the real-time operating condition parameter set at the trigger point, flutter risk prediction information output by the model, and the final vibration response data after intervention. To improve the model's prediction accuracy, the system uses the gradient descent algorithm to refine the model's internal parameters. Perform an online update. Set the model's current old parameters. The step size, or learning rate, is 0.520. The value is 0.01. The system calculates the gradient between the current prediction result and the actual response using the loss function. Assuming the calculated gradient value... The value is 0.8. Calculations based on the formula yield... The calculation results show that the model's internal parameters were updated from 0.520 to 0.512. This fine-tuning reflects the model's correction of the correlation weights between operating conditions and risks based on measured feedback. Through this structured correlation and incremental learning, the dynamic prediction model can continuously absorb operational experience, achieving self-iteration and continuous evolution of flutter risk prediction accuracy under complex and ever-changing zero-output operating conditions.

[0070] Based on the same inventive concept, such as Figure 4 As shown, the present invention also provides a system for predicting and suppressing the risk of flutter on rotating machinery blades, the system comprising: The operating condition parameter acquisition module is used to acquire the real-time operating condition parameter set of the steam turbine unit; The flutter risk prediction module is used to input the real-time operating condition parameter set into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk, and obtain flutter risk prediction information, which includes the predicted flutter risk level and the target sensitive area identifier. The control command generation module is used to generate control commands based on the flutter risk prediction information. The control commands are used to control the dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area identifier, thereby reconstructing the flow field distribution in the low-pressure cylinder. The control command execution module is used to execute the control command and collect the real-time vibration signal of the rotating mechanical blade; The vibration monitoring and closed-loop control module is used to monitor and adaptively control the steam injection parameters of the dynamic suppression channel based on the real-time vibration signal.

[0071] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0072] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for the prediction and suppression of the risk of flutter of a rotating machine blade, characterized in that, The method includes: Obtain the real-time operating parameter set of the steam turbine unit; The real-time operating condition parameter set is input into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk to obtain flutter risk prediction information, which includes the predicted flutter risk level and the target sensitive area identifier. Based on the flutter risk prediction information, a control command is generated. The control command is used to control the dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area identifier, thereby reconstructing the flow field distribution in the low-pressure cylinder. Execute the control command to collect real-time vibration signals of the rotating mechanical blades; Based on the real-time vibration signal, the steam injection parameters of the dynamic suppression channel are monitored and adaptively controlled.

2. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 1, characterized in that, The acquisition of the real-time operating parameter set of the steam turbine unit includes: Collect the steam inlet pressure, steam inlet temperature, exhaust vacuum, and rotor speed of the steam turbine unit to form macroscopic operating parameters; Based on the macroscopic operating parameters, the blade aerodynamic load distribution map is retrieved from a pre-set database that stores multiple sets of load data corresponding to different macroscopic operating parameters. The macroscopic operating parameters and the blade aerodynamic load distribution map are combined to generate a real-time operating condition parameter set.

3. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 1, characterized in that, The step of inputting the real-time operating condition parameter set into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk to obtain flutter risk prediction information includes: The real-time operating condition parameter set is matched and analyzed with the historical operating condition data in the dynamic prediction model to calculate the flutter occurrence probability value that quantifies the current flutter risk. Identify the dominant vibration mode and circumferential location of historical flutter events that are highly correlated with the real-time operating condition parameter set; By integrating the flutter occurrence probability value, the dominant vibration mode, and the circumferential location of occurrence, flutter risk prediction information is generated, which includes the predicted flutter risk level and the target sensitive area identifier.

4. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 3, characterized in that, The generation of control instructions based on the flutter risk prediction information includes: Determine whether the predicted flutter risk level exceeds a preset risk threshold for triggering active intervention; If the limit is not exceeded, a first control command is generated to control the opening of the steady-state flow channel in the dual-channel cooling steam bypass system and maintain the basic steam flow rate. If the limit is exceeded, a second control command is generated to control the opening of the steady-state flow channel and simultaneously activate the dynamic suppression flow channel. The second control command includes the initial steam flow rate and injection angle parameters set for the activated dynamic suppression flow channel.

5. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 4, characterized in that, The injection angle parameters include: Analyze the flutter risk prediction information to identify the type of the dominant vibration mode; If the dominant vibration mode is a bending mode, then the injection angle parameter is set to the direction of the steam jet that can generate radial damping force; If the dominant vibration mode is a torsional mode, then the injection angle parameter is set to the direction of the steam jet that can generate reverse aerodynamic torque.

6. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 1, characterized in that, The step of monitoring and adaptively controlling the steam injection parameters of the dynamic suppression channel based on the real-time vibration signal includes: Extract the characteristic vibration amplitude corresponding to the dominant vibration mode indicated in the flutter risk prediction information from the real-time vibration signal; Calculate the deviation between the characteristic vibration amplitude and the preset target amplitude range characterizing the safe vibration level of the blade; Based on the deviation, a steam flow correction command is generated to adjust the dynamic suppression channel.

7. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 1, characterized in that, The method further includes: After performing closed-loop fine-tuning and stabilizing the blade vibration, the final vibration response data of the blade is obtained; The real-time operating condition parameter set that triggered this protection, the flutter risk prediction information, and the final vibration response data are correlated to form a training data sample. The internal parameters of the dynamic prediction model used to establish the correlation between operating conditions and flutter risk are updated using the training data samples.

8. The method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 4, characterized in that, The generation of the second control command includes: Based on the target sensitive area identifier, select the target nozzles that cover the area from the adjustable nozzle array; Based on the predicted flutter risk level, the initial steam flow rate is allocated to the target nozzle; The injection angle parameter is set for the target nozzle based on the dominant vibration mode.

9. A method for predicting and suppressing the risk of flutter in rotating machinery blades according to claim 6, characterized in that, The step of generating a steam flow correction command for adjusting the dynamically suppressed flow channel based on the deviation includes: The characteristic vibration amplitude is compared with the upper and lower threshold values ​​of the target amplitude range; If the characteristic vibration amplitude is greater than the upper limit threshold, a first correction instruction is generated to increase the steam flow rate of the dynamic suppression channel. The first correction instruction includes the flow rate increase calculated based on the deviation magnitude. If the characteristic vibration amplitude is less than the lower threshold, a second correction instruction is generated to reduce the steam flow rate of the dynamic suppression channel. The second correction instruction includes the flow rate reduction calculated based on the deviation magnitude. If the characteristic vibration amplitude is within the target amplitude range, a third correction command is generated to maintain the current steam flow rate of the dynamic suppression channel.

10. A system for predicting and suppressing the risk of flutter in rotating machinery blades, applied to the method for predicting and suppressing the risk of flutter in rotating machinery blades as described in any one of claims 1-9, characterized in that, The system includes: The operating condition parameter acquisition module is used to acquire the real-time operating condition parameter set of the steam turbine unit; The flutter risk prediction module is used to input the real-time operating condition parameter set into a preset dynamic prediction model for establishing the correlation between operating conditions and flutter risk, and obtain flutter risk prediction information, which includes the predicted flutter risk level and the target sensitive area identifier. The control command generation module is used to generate control commands based on the flutter risk prediction information. The control commands are used to control the dual-channel cooling steam bypass system with a steady-state flow channel and a dynamic suppression flow channel to inject steam into the local flow field corresponding to the target sensitive area identifier, thereby reconstructing the flow field distribution in the low-pressure cylinder. The control command execution module is used to execute the control command and collect the real-time vibration signal of the rotating mechanical blade; The vibration monitoring and closed-loop control module is used to monitor and adaptively control the steam injection parameters of the dynamic suppression channel based on the real-time vibration signal.

Images