Intelligent early warning system and method for fluttering of last-stage blade of low-pressure cylinder

Abstract

The invention relates to the technical field of low-pressure cylinder last-stage blade monitoring, and discloses a low-pressure cylinder last-stage blade flutter intelligent early warning system and method, and the system comprises a first construction module which is used for constructing an unsteady CFD model; the second construction module is used for setting a plurality of monitoring areas and a plurality of monitoring point positions of each monitoring area, and constructing an association mapping table; the judgment module is used for acquiring real-time operation parameters and real-time static pressure pulsation data of a plurality of monitoring point positions, and judging whether a flutter risk exists or not in combination with an unsteady CFD model and an association mapping table; and the early warning module is used for determining the predicted flutter level and generating an early warning instruction if the flutter risk exists, so that the early warning accuracy and reliability are effectively improved, and a powerful guarantee is provided for safe operation of the low-pressure cylinder last-stage blade.

Description

Technical Field

[0001] This application relates to the field of low-pressure cylinder last-stage blade monitoring technology, and in particular to an intelligent early warning system and method for low-pressure cylinder last-stage blade flutter. Background Technology

[0002] As an important component of the steam turbine, the low-pressure cylinder experiences a sudden change in steam flow when the unit performs a low-pressure cylinder cut-off operation. Under the condition of a sudden decrease in steam flow, the flow field will be significantly distorted, which can easily cause blade flutter and lead to stress concentration on the blades. If it is not detected and intervened in time, it can cause fatigue fracture of the blades in a short period of time, which can lead to major safety accidents such as equipment shutdown or even damage to the unit.

[0003] In existing technologies, conventional vibration sensors are mostly used for monitoring, focusing on the overall vibration amplitude of the blade. However, there is a lack of in-depth analysis and utilization of the correlation between dynamic changes in the flow field under complex operating conditions and blade flutter, resulting in low accuracy and reliability of early warnings and missing the best time for intervention. Summary of the Invention

[0004] To address the aforementioned technical issues, this application provides an intelligent early warning system and method for flutter in the last stage blades of a low-pressure cylinder. By constructing an unsteady CFD model, setting multiple monitoring areas and monitoring points, and building an associated mapping table, combined with real-time operating parameters and monitoring data, the system can more accurately and promptly determine the existence of flutter risk, identify the predicted flutter level, and generate early warning commands. This effectively improves the accuracy and reliability of the early warning, providing strong protection for the safe operation of the last stage blades of the low-pressure cylinder.

[0005] In some embodiments of this application, a low-pressure cylinder last-stage blade flutter intelligent early warning system is provided, including:

[0006] The first building block is used to build unsteady CFD models;

[0007] The second construction module is used to set multiple monitoring areas and several monitoring points in each monitoring area, and to build an association mapping table;

[0008] The judgment module is used to obtain real-time operating parameters and real-time static pressure pulsation data of several monitoring points, and combine them with unsteady CFD models and correlation mapping tables to determine whether there is a risk of flutter.

[0009] The early warning module is used to determine the predicted flutter level and generate an early warning command if there is a risk of flutter.

[0010] In some embodiments of this application, constructing an unsteady CFD model includes:

[0011] Obtain the actual structural information of the current low-pressure cylinder last stage blades, and construct a three-dimensional geometric model based on the actual structural information;

[0012] Several operating condition characteristic indicators are set based on historical operating condition parameters;

[0013] Multiple working condition scenarios are generated based on all working condition characteristic indicators, and each working condition scenario is mapped to several working condition characteristic indicators and corresponding characteristic working condition parameters.

[0014] The characteristic operating parameters of each operating scenario are used as initial conditions, and combined with unsteady flow calculation methods, the dynamic changes of the flow field under each operating scenario are captured.

[0015] Generate the coupling relationship between the dynamic changes of the flow field and the three-dimensional geometric model under all working conditions, and construct an unsteady CFD model based on the coupling relationship.

[0016] In some embodiments of this application, multiple monitoring areas and several monitoring points are set in each monitoring area, including:

[0017] Several monitoring areas for the last stage blades of the low-pressure cylinder are pre-defined;

[0018] Obtain historical flutter logs under different operating conditions, extract historical flutter levels, several historical flutter features of each monitoring area, and historical apparent duration from the historical flutter logs, and construct a mapping table of operating conditions-flutter levels-apparent duration;

[0019] Pre-set the threshold for the explicit duration of flutter level for each working condition scenario;

[0020] Filter out several monitoring areas in the working condition scenario-flutter level-visibility duration mapping table where the historical visibility duration is greater than the corresponding preset visibility duration threshold, and set them as the visibility areas of the corresponding historical flutter level under the corresponding working condition scenario.

[0021] Generate monitoring evaluation values ​​for each monitoring area;

[0022] The number of monitoring points in the corresponding monitoring area is set according to the monitoring evaluation value, and several monitoring points are set in the monitoring area in combination with the flow field change characteristics in the monitoring area.

[0023] In some embodiments of this application, generating a monitoring evaluation value for each monitoring area includes:

[0024] Calculate the difference between the historical visibility duration and the corresponding preset visibility duration threshold for the same monitoring area that is a visible area;

[0025] Calculate the occurrence coefficient of each working condition scenario and the influence coefficient of each historical flutter level under the corresponding working condition scenario, and generate the weight coefficient of the corresponding historical flutter level based on the occurrence coefficient and the influence coefficient.

[0026] Calculate the number of visible areas with different historical flutter levels under all operating conditions within the same monitoring area;

[0027] The number of visible areas in the same monitoring area, the time difference of several visible areas, and the weight coefficients of several historical flutter levels are weighted and summed, and then the monitoring evaluation is transformed to obtain the monitoring evaluation value of the corresponding monitoring area.

[0028] In some embodiments of this application, constructing an association mapping table includes:

[0029] Randomly select a working scenario as the target working scenario;

[0030] The historical flutter characteristics of each monitoring point at different historical flutter levels under the target working condition scenario are obtained and quantified to obtain the corresponding historical quantified values.

[0031] Construct a sequence of historical quantization values ​​for each historical flutter feature at the same monitoring point;

[0032] Compare and analyze the same historical quantization value sequence of different monitoring points under the same historical flutter level, determine whether there is a correlation based on the analysis results, and if so, obtain several associated points for each monitoring point, determine the correlation type between each monitoring point and the corresponding associated points, and calculate the corresponding correlation coefficient.

[0033] The association types include previous association, same association, and subsequent association;

[0034] Based on the correlation type, determine the preceding correlation point sequence, the same correlation point sequence, and the subsequent correlation point sequence for each monitoring point under different historical flutter characteristics at the same historical flutter level;

[0035] Construct a sub-association mapping table for the target working condition scenario;

[0036] The sub-association mapping table includes a sequence of monitoring points for several historical flutter levels in the target working condition scenario, and each monitoring point in the monitoring point sequence is mapped to a previous associated point sequence, a same associated point sequence, and a subsequent associated point sequence under different historical flutter characteristics.

[0037] Generate a sub-association mapping table for each working scenario, and construct the association mapping table.

[0038] In some embodiments of this application, before determining whether there is a risk of flutter, the following steps are taken:

[0039] Obtain real-time operating parameters, and determine several real-time operating condition characteristic indicators and corresponding real-time characteristic operating condition parameters based on the real-time operating parameters.

[0040] A similarity analysis was performed between several real-time operating condition characteristic indicators and their corresponding real-time characteristic operating condition parameters and several operating condition scenarios to obtain the indicator similarity coefficients and their corresponding parameter similarity coefficients.

[0041] The comprehensive similarity coefficient is obtained by summing the weights of the index similarity coefficient, the corresponding parameter similarity coefficient, and the weight coefficient of each real-time operating condition characteristic index.

[0042] All operating scenarios are sorted according to the comprehensive similarity coefficient, and the operating scenario ranked first is set as the reference operating scenario for real-time operating parameters.

[0043] In some embodiments of this application, real-time operating parameters and real-time static pressure pulsation data from several monitoring points are obtained, and an unsteady CFD model and an association mapping table are used to determine whether there is a risk of flutter, including:

[0044] Real-time static pressure pulsation data from several monitoring points and reference operating conditions are input into an unsteady CFD model to obtain the flow field evolution prediction results for future periods.

[0045] A predicted flutter risk coefficient is generated based on a pre-set flutter risk prediction model and flow field evolution prediction results.

[0046] Pre-set flutter risk threshold;

[0047] If the predicted flutter risk coefficient is less than the flutter risk coefficient threshold, it is determined that there is no flutter risk.

[0048] If the predicted flutter risk coefficient is not less than the flutter risk coefficient threshold, it is determined that there is flutter risk and the first-level predicted flutter level is determined.

[0049] Based on the first-level predicted flutter level and the reference working condition scenario, several real-time standard flutter features are generated for each monitoring point, and the corresponding real-time standard quantization values ​​are obtained.

[0050] Extract the association mapping information of the first-level predicted flutter level under the reference working condition scenario in the association mapping table, and generate the prediction confidence coefficient of the first-level predicted flutter level based on the association mapping information.

[0051] Pre-set a confidence coefficient threshold;

[0052] If the prediction confidence coefficient is greater than the confidence coefficient threshold, the first-level predicted flutter level will not be corrected, and the first-level predicted flutter level will be used as the final predicted flutter level.

[0053] If the prediction confidence coefficient is not greater than the confidence coefficient threshold, the first-level predicted flutter level is corrected until the prediction confidence coefficient is greater than the confidence coefficient threshold, and a second-level predicted flutter level is generated and used as the final predicted flutter level.

[0054] In some embodiments of this application, generating a prediction confidence coefficient for the first-level predicted flutter level based on the association mapping information includes:

[0055] The associated mapping information includes the sequence of previous associated points, the sequence of subsequent associated points, and the sequence of points with the same associated points for each monitoring point under all historical flutter features corresponding to the first-level predicted flutter level in the reference working condition scenario.

[0056] Several real-time flutter features and corresponding real-time quantization values ​​are generated at each monitoring point, and compared with several real-time standard flutter features and corresponding real-time standard quantization values. The first prediction confidence coefficient is generated based on the comparison results.

[0057] The real-time flutter characteristics and corresponding real-time quantization values ​​of each associated point in the preceding associated point sequence, the same associated point sequence, and the subsequent associated point sequence of each monitoring point under different real-time flutter characteristics are obtained, and correlation analysis is performed with the real-time flutter characteristics and real-time quantization values ​​of the current monitoring point. A second prediction confidence coefficient is generated based on the correlation analysis results.

[0058] The prediction confidence coefficient is obtained by summing the weights of the first prediction confidence coefficient, the second prediction confidence coefficient, and the corresponding weight coefficients.

[0059] In some embodiments of this application, generating a warning instruction includes:

[0060] A first preset flutter level range, a second preset flutter level range, and a third preset flutter level range are preset.

[0061] When the predicted flutter level is within the first preset flutter level range, a level one warning command is generated.

[0062] When the predicted flutter level is within the second preset flutter level range, a level two warning command is generated;

[0063] When the predicted flutter level is within the third preset flutter level range, a level 3 warning command is generated.

[0064] In some embodiments of this application, a method for intelligent early warning of flutter in the last stage blades of a low-pressure cylinder is also included:

[0065] Constructing an unsteady CFD model;

[0066] Multiple monitoring areas and several monitoring points in each monitoring area are set up, and an association mapping table is constructed;

[0067] Acquire real-time operating parameters and real-time static pressure pulsation data from several monitoring points, and combine them with an unsteady CFD model and a correlation mapping table to determine whether there is a risk of flutter.

[0068] If flutter risk exists, determine the predicted flutter level and generate an early warning instruction.

[0069] The intelligent early warning system and method for low-pressure cylinder last-stage blade flutter according to embodiments of this application have the following advantages compared with the prior art:

[0070] By constructing an unsteady CFD model, setting multiple monitoring areas and monitoring points, and building an association mapping table, combined with real-time operating parameters and monitoring data, it is possible to more accurately and timely determine whether there is a risk of flutter, determine the predicted flutter level, and generate early warning instructions, which effectively improves the accuracy and reliability of early warning and provides a strong guarantee for the safe operation of the last stage blades of the low-pressure cylinder. Attached Figure Description

[0071] Figure 1 This is a schematic diagram of a low-pressure cylinder last-stage blade flutter intelligent early warning system in an embodiment of this application. Detailed Implementation

[0072] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are used to illustrate this application, but are not intended to limit the scope of this application.

[0073] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0074] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0075] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0076] like Figure 1 As shown in the figure, an intelligent early warning system for low-pressure cylinder last-stage blade flutter according to an embodiment of this application includes:

[0077] The first building block is used to build unsteady CFD models;

[0078] The second construction module is used to set multiple monitoring areas and several monitoring points in each monitoring area, and to build an association mapping table;

[0079] The judgment module is used to obtain real-time operating parameters and real-time static pressure pulsation data of several monitoring points, and combine them with unsteady CFD models and correlation mapping tables to determine whether there is a risk of flutter.

[0080] The early warning module is used to determine the predicted flutter level and generate an early warning command if there is a risk of flutter.

[0081] In this embodiment, each monitoring point includes a static pressure sensor and a data transmission unit. The static pressure sensor is a high-temperature resistant and vibration-resistant sensor to ensure that it can fully capture pressure fluctuations caused by changes in the flow field. The data transmission unit transmits real-time static pressure pulsation data to ensure the real-time performance and reliability of data transmission.

[0082] In this embodiment, not only is the overall vibration of the blade considered, but the correlation between the dynamic changes of the flow field and blade flutter under complex operating conditions is also analyzed in depth. This allows for a more comprehensive assessment of the blade flutter risk. By constructing an unsteady CFD model, the flow field changes under different operating conditions can be simulated, providing accurate data support for subsequent flutter risk assessment. At the same time, multiple monitoring areas and monitoring points are set, and an association mapping table is constructed, enabling the system to more accurately capture the flutter characteristics of the blade, improve the accuracy of early warning, and generate early warning instructions in a timely manner. This provides operators with sufficient intervention time and avoids major safety accidents such as equipment downtime or unit damage.

[0083] In some embodiments of this application, constructing an unsteady CFD model includes:

[0084] Obtain the actual structural information of the current low-pressure cylinder last stage blades, and construct a three-dimensional geometric model based on the actual structural information;

[0085] Several operating condition characteristic indicators are set based on historical operating condition parameters;

[0086] Multiple working condition scenarios are generated based on all working condition characteristic indicators, and each working condition scenario is mapped to several working condition characteristic indicators and corresponding characteristic working condition parameters.

[0087] The characteristic operating parameters of each operating scenario are used as initial conditions, and combined with unsteady flow calculation methods, the dynamic changes of the flow field under each operating scenario are captured.

[0088] Generate the coupling relationship between the dynamic changes of the flow field and the three-dimensional geometric model under all working conditions, and construct an unsteady CFD model based on the coupling relationship.

[0089] In this embodiment, historical operating condition parameters include parameters such as historical steam flow rate, historical valve opening, and historical extraction rate. Operating condition characteristic indicators are obtained by classifying all historical operating condition parameters into different types of operating condition characteristic sets. These sets cover multiple key factors for determining the flutter threshold of the last stage blades of the low-pressure cylinder. Operating condition characteristic indicators include cylinder cut-off purpose, steam flow state, and equipment state. Operating condition scenarios are obtained by combining different operating condition characteristic indicators, which can simulate the real operating environment of the last stage blades of the low-pressure cylinder under different operating conditions. Characteristic operating condition parameters refer to the specific operating parameter values ​​corresponding to the operating condition characteristic indicators in each operating condition scenario. By setting multiple operating condition scenarios and their corresponding characteristic operating condition parameters, it is possible to comprehensively cover various operating conditions that the last stage blades of the low-pressure cylinder may encounter, providing a foundation for constructing an accurate unsteady CFD model.

[0090] In this embodiment, the unsteady flow calculation method adopts a high-precision numerical simulation algorithm. This algorithm is based on the basic equations of fluid mechanics and transforms the continuous flow field problem into a discrete numerical calculation problem through discretization. It can accurately capture transient changes and complex flow phenomena in the flow field. The coupling relationship between the dynamic changes of the flow field and the three-dimensional geometric model refers to the introduction of dynamic change parameters of the flow field into the three-dimensional geometric model, so that the three-dimensional geometric model can dynamically reflect the actual changes of the flow field under different working conditions. This coupling relationship fully considers the interaction between the flow field and the blade structure, thereby making the constructed unsteady CFD model closer to the real working conditions.

[0091] In this embodiment, an unsteady CFD model is obtained by constructing a three-dimensional geometric model and determining the coupling relationship with the dynamic changes of the flow field under different operating conditions. This model can accurately simulate the flow field characteristics of the last stage blade of the low-pressure cylinder under different operating conditions, providing a reliable basis for subsequent flutter risk assessment.

[0092] In some embodiments of this application, multiple monitoring areas and several monitoring points are set in each monitoring area, including:

[0093] Several monitoring areas for the last stage blades of the low-pressure cylinder are pre-defined;

[0094] Obtain historical flutter logs under different operating conditions, extract historical flutter levels, several historical flutter features of each monitoring area, and historical apparent duration from the historical flutter logs, and construct a mapping table of operating conditions-flutter levels-apparent duration;

[0095] Pre-set the threshold for the explicit duration of flutter level for each working condition scenario;

[0096] Filter out several monitoring areas in the working condition scenario-flutter level-visibility duration mapping table where the historical visibility duration is greater than the corresponding preset visibility duration threshold, and set them as the visibility areas of the corresponding historical flutter level under the corresponding working condition scenario.

[0097] Generate monitoring evaluation values ​​for each monitoring area;

[0098] The number of monitoring points in the corresponding monitoring area is set according to the monitoring evaluation value, and several monitoring points are set in the monitoring area in combination with the flow field change characteristics in the monitoring area.

[0099] In this embodiment, the pre-defined monitoring area covers the entire last stage blade of the low-pressure cylinder. However, the area corresponding to the monitoring area is different and is set according to whether the flutter phenomenon is more likely to occur. The monitoring area is smaller for areas where flutter phenomenon is more likely to occur, and larger for areas where flutter phenomenon is less likely to occur.

[0100] In this embodiment, the historical flutter log includes several historical flutter features and corresponding historical flutter level information, which are obtained by converting historical static pressure pulsation data at multiple historical time points in each monitoring area under different operating conditions.

[0101] In this embodiment, historical flutter features refer to the corresponding flutter features that can accurately predict the corresponding historical flutter level, and historical manifestation duration refers to the interval between the historical flutter features when the corresponding historical flutter level is predicted and the actual occurrence of the corresponding historical flutter level. For example, when the historical flutter level is level one, the corresponding historical flutter feature may be that the blade vibration frequency reaches a certain specific value, and the historical manifestation duration is the time interval between the time point when the specific value occurs and the time point when the historical flutter level is determined.

[0102] In this embodiment, the preset explicit duration threshold refers to the minimum interval duration at which the flutter feature predicts the corresponding flutter level. It is set according to the influence degree of the corresponding historical flutter level and the frequency of occurrence of the historical flutter level. If a certain historical flutter level has a large influence degree and a high frequency of occurrence, its preset explicit duration threshold is relatively long, and vice versa.

[0103] In this embodiment, by acquiring historical flutter logs and constructing a mapping table of working conditions, flutter levels, and manifest duration, and by combining a preset manifest duration threshold to filter out manifest areas, the monitoring evaluation value of each monitoring area is calculated and monitoring points are reasonably arranged. This enables accurate assessment of the early warning performance and accuracy of each monitoring area for predicting flutter risks, providing strong support for subsequent assessment and early warning of flutter risks.

[0104] In some embodiments of this application, generating a monitoring evaluation value for each monitoring area includes:

[0105] Calculate the difference between the historical visibility duration and the corresponding preset visibility duration threshold for the same monitoring area that is a visible area;

[0106] Calculate the occurrence coefficient of each working condition scenario and the influence coefficient of each historical flutter level under the corresponding working condition scenario, and generate the weight coefficient of the corresponding historical flutter level based on the occurrence coefficient and the influence coefficient.

[0107] Calculate the number of visible areas with different historical flutter levels under all operating conditions within the same monitoring area;

[0108] The number of visible areas in the same monitoring area, the time difference of several visible areas, and the weight coefficients of several historical flutter levels are weighted and summed, and then the monitoring evaluation is transformed to obtain the monitoring evaluation value of the corresponding monitoring area.

[0109] In this embodiment, Where n is the number of visible regions, Δti is the duration difference when the current monitoring region is the i-th visible region, qi is the weighting coefficient of the historical flutter level when the current monitoring region is the i-th visible region, and r is the monitoring evaluation conversion coefficient.

[0110] In this embodiment, the monitoring and evaluation conversion coefficient refers to converting the sum of duration differences into a value with the same dimension as the monitoring and evaluation value. When the sum of duration differences is larger, the corresponding monitoring and evaluation value is larger, and vice versa.

[0111] In this embodiment, the occurrence coefficient refers to the ratio of the frequency of occurrence of a particular operating condition scenario to the total number of operating condition scenarios, reflecting the prevalence of that scenario in overall operation. The influence coefficient, on the other hand, is set based on historical data, considering the impact of historical flutter levels on the operational safety of the low-pressure cylinder's last stage blades under specific operating conditions. A higher influence coefficient is generated when both the degree of influence and the probability of occurrence are greater. By combining the occurrence coefficient and the influence coefficient, we can assign a reasonable weighting coefficient to each historical flutter level to reflect its importance under different operating conditions.

[0112] In this embodiment, the number of monitoring points for each monitoring area is determined by calculating the monitoring evaluation value. Areas with higher monitoring evaluation values ​​indicate stronger and more accurate early warning performance for predicting flutter risks, thus requiring more monitoring points to capture more detailed flow field change information. Conversely, areas with lower monitoring evaluation values ​​can have fewer monitoring points to optimize resource allocation, ensuring that the monitoring points within each monitoring area can fully cover key areas of flow field change while avoiding resource waste and redundancy.

[0113] In some embodiments of this application, constructing an association mapping table includes:

[0114] Randomly select a working scenario as the target working scenario;

[0115] The historical flutter characteristics of each monitoring point at different historical flutter levels under the target working condition scenario are obtained and quantified to obtain the corresponding historical quantified values.

[0116] Construct a sequence of historical quantization values ​​for each historical flutter feature at the same monitoring point;

[0117] Compare and analyze the same historical quantization value sequence of different monitoring points under the same historical flutter level, determine whether there is a correlation based on the analysis results, and if so, obtain several associated points for each monitoring point, determine the correlation type between each monitoring point and the corresponding associated points, and calculate the corresponding correlation coefficient.

[0118] The association types include previous association, same association, and subsequent association;

[0119] Based on the correlation type, determine the preceding correlation point sequence, the same correlation point sequence, and the subsequent correlation point sequence for each monitoring point under different historical flutter characteristics at the same historical flutter level;

[0120] Construct a sub-association mapping table for the target working condition scenario;

[0121] The sub-association mapping table includes a sequence of monitoring points for several historical flutter levels in the target working condition scenario, and each monitoring point in the monitoring point sequence is mapped to a previous associated point sequence, a same associated point sequence, and a subsequent associated point sequence under different historical flutter characteristics.

[0122] Generate a sub-association mapping table for each working scenario, and construct the association mapping table.

[0123] In this embodiment, quantifying historical flutter features refers to converting these features into quantifiable and comparable numerical forms. Specifically, for each historical flutter feature, an appropriate quantization method is selected based on its characteristics. For example, for features related to vibration frequency, the frequency value can be directly used as the quantization value; for features related to vibration amplitude, a certain amplitude range can be set, and different quantization value intervals can be used for quantization. Through quantization, each historical flutter feature is assigned a specific quantization value, which can more intuitively reflect the magnitude and changes of the historical flutter features.

[0124] In this embodiment, continuous historical time nodes refer to time points selected at regular intervals within the period 150 seconds before the historical flutter level determination node. These time points can continuously record the changes in historical flutter characteristics at each monitoring point, providing continuous and detailed data support for subsequent analysis.

[0125] In this embodiment, when determining whether there is a correlation between sequences, the main focus is on the similarity or difference between the sequences. If there is similarity, that is, the feature quantification value sequence of a certain monitoring point is similar to the sequence of another monitoring point in terms of change trend or value, then it is determined that there is a correlation between the two monitoring points. Otherwise, if there is a difference, then there is no correlation.

[0126] In this embodiment, the association type is determined by the time sequence of the association relationship. That is, the previous association is that the historical quantization value change of another monitoring point is earlier than and greater than the historical quantization value of the current monitoring point; the same association is that the historical quantization value change of another monitoring point and the current monitoring point occur simultaneously and are equal; and the subsequent association is that the feature quantization value change of another monitoring point lags behind and is less than the historical quantization value of the current monitoring point.

[0127] In this embodiment, the correlation coefficient can be calculated based on statistical indicators such as correlation coefficient and similarity. These indicators can accurately measure the degree of correlation between the feature quantification value sequences of different monitoring points. The correlation coefficient usually ranges from 0 to 1. The larger the value, the stronger the correlation coefficient, and vice versa.

[0128] In this embodiment, the previous associated point sequence, the same associated point sequence, and the subsequent associated point sequence are all constructed by sorting the associated points under the corresponding historical flutter characteristics and the current monitoring points from largest to smallest.

[0129] In this embodiment, the monitoring point sequence is sorted according to the sub-monitoring evaluation values ​​and flow field change characteristics of the monitoring areas under different historical flutter levels in the target operating condition scenario. The sub-monitoring evaluation values ​​are calculated based on the number of visible areas, duration difference, and weighting coefficient of the monitoring area under all historical flutter levels in the target operating condition scenario. The monitoring areas are sorted by the sub-monitoring evaluation values. The higher the ranking of the monitoring area, the higher the ranking of the monitoring point in the corresponding monitoring area in the monitoring point sequence. At the same time, combined with the flow field change characteristics, it is ensured that the sorting results can accurately reflect the importance and sensitivity of the monitoring points for flutter risk prediction in the target operating condition scenario.

[0130] In this embodiment, the sub-association mapping table records in detail the correlation between monitoring points of different historical flutter characteristics at different historical flutter levels under each operating scenario. This provides strong data support for the entire low-pressure cylinder last-stage blade flutter intelligent early warning system, enabling the system to comprehensively consider the correlation between various monitoring points when judging flutter risk, evaluate the prediction accuracy and reliability of the unsteady CFD model, improve the accuracy of flutter risk prediction, and take corresponding preventive measures in advance to ensure the safe and stable operation of the low-pressure cylinder last-stage blade.

[0131] In some embodiments of this application, before determining whether there is a risk of flutter, the following steps are taken:

[0132] Obtain real-time operating parameters, and determine several real-time operating condition characteristic indicators and corresponding real-time characteristic operating condition parameters based on the real-time operating parameters.

[0133] A similarity analysis was performed between several real-time operating condition characteristic indicators and their corresponding real-time characteristic operating condition parameters and several operating condition scenarios to obtain the indicator similarity coefficients and their corresponding parameter similarity coefficients.

[0134] The comprehensive similarity coefficient is obtained by summing the weights of the index similarity coefficient, the corresponding parameter similarity coefficient, and the weight coefficient of each real-time operating condition characteristic index.

[0135] All operating scenarios are sorted according to the comprehensive similarity coefficient, and the operating scenario ranked first is set as the reference operating scenario for real-time operating parameters.

[0136] In this embodiment, the index similarity coefficient refers to the degree of matching between the real-time operating condition characteristic index and the operating condition characteristic index of each operating condition scenario, and the parameter similarity coefficient refers to the degree of closeness between the real-time characteristic operating condition parameter and the characteristic operating condition parameter. When the degree of matching is higher and the degree of closeness is higher, the corresponding index similarity coefficient is larger and the parameter similarity coefficient is larger, that is, the comprehensive similarity coefficient is larger, and vice versa.

[0137] In this embodiment, the operating scenarios are sorted from largest to smallest according to the comprehensive similarity coefficient. The operating scenario with the highest comprehensive similarity coefficient and the highest similarity coefficient is set as the reference operating scenario for the real-time operating parameters, which lays the foundation for subsequent judgment on whether there is a risk of flutter. The preset similarity coefficient threshold is the minimum threshold used to judge whether the similarity between the real-time operating parameters and the operating scenario reaches a sufficient standard. This threshold is set according to actual operating needs and historical data experience.

[0138] In some embodiments of this application, real-time operating parameters and real-time static pressure pulsation data from several monitoring points are obtained, and an unsteady CFD model and an association mapping table are used to determine whether there is a risk of flutter, including:

[0139] Real-time static pressure pulsation data from several monitoring points and reference operating conditions are input into an unsteady CFD model to obtain the flow field evolution prediction results for future periods.

[0140] A predicted flutter risk coefficient is generated based on a pre-set flutter risk prediction model and flow field evolution prediction results.

[0141] Pre-set flutter risk threshold;

[0142] If the predicted flutter risk coefficient is less than the flutter risk coefficient threshold, it is determined that there is no flutter risk.

[0143] If the predicted flutter risk coefficient is not less than the flutter risk coefficient threshold, it is determined that there is flutter risk and the first-level predicted flutter level is determined.

[0144] Based on the first-level predicted flutter level and the reference working condition scenario, several real-time standard flutter features are generated for each monitoring point, and the corresponding real-time standard quantization values ​​are obtained.

[0145] Extract the association mapping information of the first-level predicted flutter level under the reference working condition scenario in the association mapping table, and generate the prediction confidence coefficient of the first-level predicted flutter level based on the association mapping information.

[0146] Pre-set a confidence coefficient threshold;

[0147] If the prediction confidence coefficient is greater than the confidence coefficient threshold, the first-level predicted flutter level will not be corrected, and the first-level predicted flutter level will be used as the final predicted flutter level.

[0148] If the prediction confidence coefficient is not greater than the confidence coefficient threshold, the first-level predicted flutter level is corrected until the prediction confidence coefficient is greater than the confidence coefficient threshold, and a second-level predicted flutter level is generated and used as the final predicted flutter level.

[0149] In this embodiment, the preset flutter risk prediction model is a mathematical model constructed based on a large amount of historical flutter data and the corresponding flow field evolution characteristics. This model can comprehensively consider the influence of multiple factors on flutter risk, and can accurately generate the predicted flutter risk coefficient by inputting the flow field evolution prediction results.

[0150] In this embodiment, the predicted flutter level is determined based on a pre-constructed flutter risk coefficient-flutter level mapping table. This mapping table records in detail the flutter levels corresponding to different ranges of predicted flutter risk coefficients. By finding the position of the predicted flutter risk coefficient in the mapping table, the first-level predicted flutter level can be determined quickly and accurately.

[0151] In this embodiment, the real-time standard flutter feature refers to the standard flutter feature that each monitoring point should exhibit under the first-level predicted flutter level and reference working condition scenario. These features are quantized to obtain the corresponding real-time standard quantized value, which is used to compare and analyze with the actual monitored flutter feature. The prediction confidence coefficient is calculated based on the analysis results and the associated mapping information.

[0152] In this embodiment, the prediction confidence coefficient is an indicator used to evaluate the accuracy of the first-level predicted flutter level. The confidence coefficient threshold is the minimum confidence coefficient required for the accuracy of the first-level predicted flutter level to meet the requirements. It is set based on historical data. When the prediction confidence coefficient is greater than the confidence coefficient threshold, it indicates that the first-level predicted flutter level has high accuracy and does not need to be corrected. Conversely, the first-level predicted flutter level needs to be corrected to improve the prediction accuracy.

[0153] In this embodiment, correcting the first-level predicted flutter level refers to adjusting the predicted flutter level according to the correlation mapping information and the degree of difference between the actual monitored flutter characteristics and the real-time standard flutter characteristics, based on preset correction rules and algorithms. The adjusted predicted flutter level is the second-level predicted flutter level. The preset correction rules and algorithms are formulated based on correction experience and patterns in similar situations from historical data. They can reasonably correct the first-level predicted flutter level according to different degrees of difference and correlation mapping information, ensuring that the second-level predicted flutter level is more accurate and reliable. Through this correction mechanism, the accuracy and reliability of the entire low-pressure cylinder last-stage blade flutter intelligent early warning system for predicting flutter risks can be further improved, providing stronger protection for the safe and stable operation of the low-pressure cylinder last-stage blades.

[0154] In this embodiment, the secondary predicted flutter level is a more accurate prediction result obtained after correcting the primary predicted flutter level. It can more accurately reflect the actual flutter situation of the last stage blades of the low-pressure cylinder in the future period, providing strong support for taking corresponding preventive measures.

[0155] In some embodiments of this application, generating a prediction confidence coefficient for the first-level predicted flutter level based on the association mapping information includes:

[0156] The associated mapping information includes the sequence of previous associated points, the sequence of subsequent associated points, and the sequence of points with the same associated points for each monitoring point under all historical flutter features corresponding to the first-level predicted flutter level in the reference working condition scenario.

[0157] Several real-time flutter features and corresponding real-time quantization values ​​are generated at each monitoring point, and compared with several real-time standard flutter features and corresponding real-time standard quantization values. The first prediction confidence coefficient is generated based on the comparison results.

[0158] The real-time flutter characteristics and corresponding real-time quantization values ​​of each associated point in the preceding associated point sequence, the same associated point sequence, and the subsequent associated point sequence of each monitoring point under different real-time flutter characteristics are obtained, and correlation analysis is performed with the real-time flutter characteristics and real-time quantization values ​​of the current monitoring point. A second prediction confidence coefficient is generated based on the correlation analysis results.

[0159] The prediction confidence coefficient is obtained by summing the weights of the first prediction confidence coefficient, the second prediction confidence coefficient, and the corresponding weight coefficients.

[0160] In this embodiment, pre-set pre-association conditions, same-association conditions, and post-association conditions are defined. The pre-association conditions include several first standard quantization value difference intervals between the monitoring point and each associated point under different real-time flutter characteristics. The same-association conditions include several second standard quantization value difference intervals between the monitoring point and each associated point under different real-time flutter characteristics. The post-association conditions include several third standard quantization value difference intervals between the monitoring point and each associated point under different real-time flutter characteristics. The standard quantization value difference intervals are set based on the correlation coefficient between the associated point and the monitoring point under different real-time flutter characteristics and the historical quantization value differences.

[0161] In this embodiment, the difference between the real-time quantization value corresponding to the real-time flutter feature at each associated point under each real-time standard flutter feature and the real-time quantization value corresponding to the real-time flutter feature at the current monitoring point is calculated. If the number of real-time quantization value differences falls within the corresponding standard quantization value difference interval, the corresponding association condition is satisfied. If the number of associated points satisfying the association condition reaches a preset proportion threshold of the total number of associated points in the previous associated point sequence, then based on the ratio of the number of associated points satisfying the association condition to the preset proportion threshold, combined with preset weights, an association satisfaction coefficient is generated (the association satisfaction coefficient includes the previous association satisfaction coefficient, the same association satisfaction coefficient, and the subsequent association satisfaction coefficient, with corresponding weight coefficients of 0.2, 0.3, and 0.5, respectively). The previous association satisfaction coefficient, the same association satisfaction coefficient, and the subsequent association satisfaction coefficient are weighted and summed to obtain the second prediction confidence coefficient corresponding to the association analysis result. When the association satisfaction coefficient is larger, the corresponding second prediction confidence coefficient is larger.

[0162] In this embodiment, the weighting coefficients of the first prediction confidence coefficient and the second prediction confidence coefficient are set based on historical data. In this application, the weighting coefficient of the first prediction confidence coefficient is 0.4, and the weighting coefficient of the second prediction confidence coefficient is 0.6. Through weighting, the first prediction confidence coefficient and the second prediction confidence coefficient are merged into a comprehensive prediction confidence coefficient, which can more comprehensively and accurately assess the accuracy of the first-level predicted flutter level.

[0163] In this embodiment, the generation of prediction confidence coefficients based on correlation mapping information, the determination of whether to correct the first-level predicted flutter level, and the formulation of reasonable correction strategies provide a more scientific and reliable evaluation method for the intelligent early warning system of flutter in the last stage blades of low-pressure cylinders.

[0164] In some embodiments of this application, generating a warning instruction includes:

[0165] A first preset flutter level range, a second preset flutter level range, and a third preset flutter level range are preset.

[0166] When the predicted flutter level is within the first preset flutter level range, a level one warning command is generated.

[0167] When the predicted flutter level is within the second preset flutter level range, a level two warning command is generated;

[0168] When the predicted flutter level is within the third preset flutter level range, a level 3 warning command is generated.

[0169] In this embodiment, the preset flutter level range is set based on historical flutter logs, and the first preset flutter level range < the second preset flutter level range < the third preset flutter level range.

[0170] In this embodiment, each warning command corresponds to a specific preventive measure. The preventive measure for a Level 1 warning command is to increase the monitoring frequency, closely monitor the operating status of the last stage blades of the low-pressure cylinder, and record relevant data for subsequent analysis. The preventive measure for a Level 2 warning command is to appropriately adjust the operating parameters of the last stage blades of the low-pressure cylinder, such as adjusting parameters like steam flow and pressure, while also checking whether the blade fixing devices are loose. The preventive measure for a Level 3 warning command is to immediately stop the machine for inspection and conduct a comprehensive and detailed inspection of the last stage blades of the low-pressure cylinder.

[0171] In this embodiment, through this graded early warning and corresponding preventive measures, effective measures can be taken in a timely manner according to different flutter levels, so as to maximize the safe and stable operation of the last stage blades of the low-pressure cylinder and avoid serious consequences such as equipment damage caused by flutter problems.

[0172] In some embodiments of this application, a method for intelligent early warning of flutter in the last stage blades of a low-pressure cylinder is also included:

[0173] Constructing an unsteady CFD model;

[0174] Multiple monitoring areas and several monitoring points in each monitoring area are set up, and an association mapping table is constructed;

[0175] Acquire real-time operating parameters and real-time static pressure pulsation data from several monitoring points, and combine them with an unsteady CFD model and a correlation mapping table to determine whether there is a risk of flutter.

[0176] If flutter risk exists, determine the predicted flutter level and generate an early warning instruction.

[0177] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of this application, and these improvements and substitutions should also be considered within the scope of protection of this application.

Claims

1. A smart early warning system for flutter in the last stage blades of a low-pressure cylinder, characterized in that, include: The first building block is used to build unsteady CFD models; The second construction module is used to set multiple monitoring areas and several monitoring points in each monitoring area, and to build an association mapping table; The judgment module is used to obtain real-time operating parameters and real-time static pressure pulsation data of several monitoring points, and combine them with unsteady CFD models and correlation mapping tables to determine whether there is a risk of flutter. The early warning module is used to determine the predicted flutter level and generate an early warning command if there is a risk of flutter.

2. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 1, characterized in that, Constructing an unsteady CFD model includes: Obtain the actual structural information of the current low-pressure cylinder last stage blades, and construct a three-dimensional geometric model based on the actual structural information; Several operating condition characteristic indicators are set based on historical operating condition parameters; Multiple working condition scenarios are generated based on all working condition characteristic indicators, and each working condition scenario is mapped to several working condition characteristic indicators and corresponding characteristic working condition parameters. The characteristic operating parameters of each operating scenario are used as initial conditions, and combined with unsteady flow calculation methods, the dynamic changes of the flow field under each operating scenario are captured. Generate the coupling relationship between the dynamic changes of the flow field and the three-dimensional geometric model under all working conditions, and construct an unsteady CFD model based on the coupling relationship.

3. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 2, characterized in that, Multiple monitoring areas and several monitoring points in each monitoring area are set up, including: Several monitoring areas for the last stage blades of the low-pressure cylinder are pre-defined; Obtain historical flutter logs under different operating conditions, extract historical flutter levels, several historical flutter features of each monitoring area, and historical apparent duration from the historical flutter logs, and construct a mapping table of operating conditions-flutter levels-apparent duration; Pre-set the threshold for the explicit duration of flutter level for each working condition scenario; Filter out several monitoring areas in the working condition scenario-flutter level-visibility duration mapping table where the historical visibility duration is greater than the corresponding preset visibility duration threshold, and set them as the visibility areas of the corresponding historical flutter level under the corresponding working condition scenario. Generate monitoring evaluation values ​​for each monitoring area; The number of monitoring points in the corresponding monitoring area is set according to the monitoring evaluation value, and several monitoring points are set in the monitoring area in combination with the flow field change characteristics in the monitoring area.

4. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 3, characterized in that, Generate monitoring evaluation values ​​for each monitoring area, including: Calculate the difference between the historical visibility duration and the corresponding preset visibility duration threshold for the same monitoring area that is a visible area; Calculate the occurrence coefficient of each working condition scenario and the influence coefficient of each historical flutter level under the corresponding working condition scenario, and generate the weight coefficient of the corresponding historical flutter level based on the occurrence coefficient and the influence coefficient. Calculate the number of visible areas with different historical flutter levels under all operating conditions within the same monitoring area; The number of visible areas in the same monitoring area, the time difference of several visible areas, and the weight coefficients of several historical flutter levels are weighted and summed, and then the monitoring evaluation is transformed to obtain the monitoring evaluation value of the corresponding monitoring area.

5. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 4, characterized in that, Construct an association mapping table, including: Randomly select a working scenario as the target working scenario; The historical flutter characteristics of each monitoring point at different historical flutter levels under the target working condition scenario are obtained and quantified to obtain the corresponding historical quantified values. Construct a sequence of historical quantization values ​​for each historical flutter feature at the same monitoring point; Compare and analyze the same historical quantization value sequence of different monitoring points under the same historical flutter level, determine whether there is a correlation based on the analysis results, and if so, obtain several associated points for each monitoring point, determine the correlation type between each monitoring point and the corresponding associated points, and calculate the corresponding correlation coefficient. The association types include previous association, same association, and subsequent association; Based on the correlation type, determine the preceding correlation point sequence, the same correlation point sequence, and the subsequent correlation point sequence for each monitoring point under different historical flutter characteristics at the same historical flutter level; Construct a sub-association mapping table for the target working condition scenario; The sub-association mapping table includes a sequence of monitoring points for several historical flutter levels in the target working condition scenario, and each monitoring point in the monitoring point sequence is mapped to a previous associated point sequence, a same associated point sequence, and a subsequent associated point sequence under different historical flutter characteristics. Generate a sub-association mapping table for each working scenario, and construct the association mapping table.

6. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 5, characterized in that, Before determining whether there is a risk of flutter, the following should be included: Obtain real-time operating parameters, and determine several real-time operating condition characteristic indicators and corresponding real-time characteristic operating condition parameters based on the real-time operating parameters. A similarity analysis was performed between several real-time operating condition characteristic indicators and their corresponding real-time characteristic operating condition parameters and several operating condition scenarios to obtain the indicator similarity coefficients and their corresponding parameter similarity coefficients. The comprehensive similarity coefficient is obtained by summing the weights of the index similarity coefficient, the corresponding parameter similarity coefficient, and the weight coefficient of each real-time operating condition characteristic index. All operating scenarios are sorted according to the comprehensive similarity coefficient, and the operating scenario ranked first is set as the reference operating scenario for real-time operating parameters.

7. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 6, characterized in that, Acquire real-time operating parameters and real-time static pressure pulsation data from several monitoring points, and combine these with an unsteady CFD model and correlation mapping table to determine whether there is a risk of flutter, including: Real-time static pressure pulsation data from several monitoring points and reference operating conditions are input into an unsteady CFD model to obtain the flow field evolution prediction results for future periods. A predicted flutter risk coefficient is generated based on a pre-set flutter risk prediction model and flow field evolution prediction results. Pre-set flutter risk threshold; If the predicted flutter risk coefficient is less than the flutter risk coefficient threshold, it is determined that there is no flutter risk. If the predicted flutter risk coefficient is not less than the flutter risk coefficient threshold, it is determined that there is flutter risk and the first-level predicted flutter level is determined. Based on the first-level predicted flutter level and the reference working condition scenario, several real-time standard flutter features are generated for each monitoring point, and the corresponding real-time standard quantization values ​​are obtained. Extract the association mapping information of the first-level predicted flutter level under the reference working condition scenario in the association mapping table, and generate the prediction confidence coefficient of the first-level predicted flutter level based on the association mapping information. Pre-set a confidence coefficient threshold; If the prediction confidence coefficient is greater than the confidence coefficient threshold, the first-level predicted flutter level will not be corrected, and the first-level predicted flutter level will be used as the final predicted flutter level. If the prediction confidence coefficient is not greater than the confidence coefficient threshold, the first-level predicted flutter level is corrected until the prediction confidence coefficient is greater than the confidence coefficient threshold, and a second-level predicted flutter level is generated and used as the final predicted flutter level.

8. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 7, characterized in that, The prediction confidence coefficient for the first-level predicted flutter level is generated based on the association mapping information, including: The associated mapping information includes the sequence of previous associated points, the sequence of subsequent associated points, and the sequence of points with the same associated points for each monitoring point under all historical flutter features corresponding to the first-level predicted flutter level in the reference working condition scenario. Several real-time flutter features and corresponding real-time quantization values ​​are generated at each monitoring point, and compared with several real-time standard flutter features and corresponding real-time standard quantization values. The first prediction confidence coefficient is generated based on the comparison results. The real-time flutter characteristics and corresponding real-time quantization values ​​of each associated point in the preceding associated point sequence, the same associated point sequence, and the subsequent associated point sequence of each monitoring point under different real-time flutter characteristics are obtained, and correlation analysis is performed with the real-time flutter characteristics and real-time quantization values ​​of the current monitoring point. A second prediction confidence coefficient is generated based on the correlation analysis results. The prediction confidence coefficient is obtained by summing the weights of the first prediction confidence coefficient, the second prediction confidence coefficient, and the corresponding weight coefficients.

9. The intelligent early warning system for low-pressure cylinder last-stage blade flutter as described in claim 8, characterized in that, Generate early warning instructions, including: A first preset flutter level range, a second preset flutter level range, and a third preset flutter level range are preset. When the predicted flutter level is within the first preset flutter level range, a level one warning command is generated. When the predicted flutter level is within the second preset flutter level range, a level two warning command is generated; When the predicted flutter level is within the third preset flutter level range, a level 3 warning command is generated.

10. A method for intelligent early warning of flutter in the last stage blades of a low-pressure cylinder, characterized in that, include: Constructing an unsteady CFD model; Multiple monitoring areas and several monitoring points in each monitoring area are set up, and an association mapping table is constructed; Acquire real-time operating parameters and real-time static pressure pulsation data from several monitoring points, and combine them with an unsteady CFD model and a correlation mapping table to determine whether there is a risk of flutter. If flutter risk exists, determine the predicted flutter level and generate an early warning instruction.

Images