Analysis method for turbine low-pressure cylinder last-stage blade fault

Abstract

The invention discloses a turbine low-pressure cylinder last-stage blade fault analysis method comprising the steps that a blade structure map is established, and a plurality of regions of interest are calibrated in the blade structure map; historical fault information of each concerned area is obtained based on the real-time operation condition of the steam turbine, the historical fault information is analyzed, and fault areas and fault features of the concerned areas are determined; adjusting the corresponding attention areas based on the fault areas and the fault features to obtain corrected attention areas, and constructing a fault association model of each corrected attention area according to a fault association relationship between the corrected attention areas; and acquiring real-time blade data of each corrected attention area, performing state evaluation on the corresponding corrected attention area based on the real-time blade data and the fault correlation model, and judging whether an alarm signal is generated or not according to a state evaluation result, thereby realizing accurate judgment on the blade state.

Description

Technical Field

[0001] The present application relates to the technical field of low-pressure cylinder blades, and in particular to a method for analyzing faults of last-stage blades of a low-pressure cylinder of a steam turbine. Background Art

[0002] The development of thermal power flexibility technology has made the operating conditions of large thermal power turbines more complicated, which has put forward higher requirements on the safety of turbine blades, especially the large flow area of ​​the low-pressure cylinder, the more drastic changes in the characteristics of the flowing medium, and even a non-steady-state flow state, which is prone to flutter and blast heating problems. The safety status of the last two stages of blades in the low-pressure cylinder is worthy of study. Therefore, in order to accurately judge the status of the last two stages of blades in the low-pressure cylinder of the turbine, an analysis method for the failure of the last stage blades of the low-pressure cylinder of the turbine is urgently needed. Summary of the invention

[0003] In order to solve the above technical problems, the present application provides an analysis method for the fault of the last stage blade of the low-pressure cylinder of a turbine. By establishing a blade structure map to calibrate multiple areas of interest, the fault area and fault characteristics of each area of ​​interest under real-time operating conditions are obtained, the corrected area of ​​interest is obtained, and a fault association model of the corrected area of ​​interest is constructed. The state evaluation is performed according to the real-time blade data and the fault association model to achieve accurate judgment of the blade state.

[0004] In some embodiments of the present application, a method for analyzing a fault of a last-stage blade of a low-pressure cylinder of a steam turbine is provided, comprising:

[0005] Establishing a leaf structure atlas, wherein a plurality of regions of interest are marked in the leaf structure atlas;

[0006] Based on the real-time operating conditions of the steam turbine, historical fault information of each area of ​​concern is obtained, the historical fault information is analyzed, and the fault area and fault characteristics of the area of ​​concern are determined;

[0007] Adjust the corresponding focus area based on the fault area and fault characteristics to obtain the modified focus area, and build a fault association model for each modified focus area according to the fault association relationship between the modified focus areas;

[0008] The real-time blade data of each modified focus area is obtained, and the status of the corresponding modified focus area is evaluated based on the real-time blade data and the fault association model, and it is determined whether to generate an alarm signal according to the status evaluation result.

[0009] In some embodiments of the present application, establishing a blade structure atlas includes:

[0010] Determine the model and real-time operating conditions of the steam turbine, and obtain the actual geometric parameters of the last-stage blades of the low-pressure cylinder of the steam turbine;

[0011] Construct the initial structural model of the last stage blade of the low pressure cylinder of the steam turbine according to the actual geometric parameters;

[0012] Evenly segment the direction from the root of the blade to one end of the blade in the initial structural model to obtain multiple focus areas in the initial structural model;

[0013] A reference coordinate system is established based on the aerodynamic center of the last stage blade of the low-pressure cylinder of the steam turbine under real-time operating conditions, and the vector coordinates of the critical position of each area of ​​interest are calibrated in the reference coordinate system;

[0014] The vector coordinates of the critical positions of different focus areas of the initial structural model are scaled proportionally to generate a blade structure map of the last-stage blade of the low-pressure cylinder of the steam turbine, and the corresponding focus areas are calibrated in the blade structure map.

[0015] In some embodiments of the present application, historical fault information is analyzed to determine the fault area and fault characteristics of the concerned area, including:

[0016] Obtain historical fault logs of the steam turbine, and classify the historical fault logs according to historical operating conditions corresponding to the historical fault logs to obtain historical fault logs of multiple categories;

[0017] According to the similarity between the real-time operating condition and the historical operating condition, the historical fault logs of the corresponding category are screened out, and the screened historical fault logs are analyzed to obtain the historical fault information of each concerned area of ​​the real-time operating condition, wherein the historical fault information includes the fault type, fault point, fault time node, fault data, fault degree and fault impact coefficient;

[0018] Determine the standard fault duration of the corresponding fault type in the corresponding area of ​​interest according to the fault time nodes in different historical fault information of the same fault type in each area of ​​interest;

[0019] Determine the fault area of ​​the corresponding fault type in the corresponding focus area according to the fault points of the same fault type in each focus area within multiple standard fault durations;

[0020] Establish a first time reference line according to the standard fault duration of the same fault type in each area of ​​concern, set a first acquisition node based on a first preset time interval, obtain fault data and fault degree within the same standard fault duration according to each first acquisition node, and map them to the first time reference line;

[0021] According to the degree of synchronous change of fault data and fault degree, characteristic fault data of corresponding fault type of each concerned area is screened out;

[0022] The standard fault durations, characteristic fault data and fault impact coefficients of multiple fault types in each area of ​​interest are set as the fault characteristics of the corresponding area of ​​interest.

[0023] In some embodiments of the present application, the corresponding region of interest is modified based on the fault region and the fault characteristics to obtain a modified region of interest, including:

[0024] Comprehensively analyze the fault characteristics of different fault types in the fault area and generate a comprehensive fault evaluation value for the corresponding fault area;

[0025] According to the relationship between the comprehensive fault evaluation value and the preset fault evaluation value interval, a correction coefficient corresponding to the fault area is set, and a corrected fault area is generated;

[0026] Determine the area ratio of the corrected fault area to the corresponding area of ​​interest based on the positional relationship between the corrected fault area and the corresponding area of ​​interest;

[0027] If the area ratio is less than the first preset area ratio, setting the corrected fault area as a corrected focus area of ​​the corresponding focus area;

[0028] If the area ratio is greater than the first preset area ratio, determine whether the corrected fault area is entirely within the corresponding focus area; if so, do not adjust the corresponding focus area; if not, obtain the area areas of other focus areas where the corrected fault area is located, and calculate the distance threshold between the area boundary points of the other focus areas where the corrected fault area is located and the boundary points of the corrected fault areas of other focus areas;

[0029] If the area of ​​the region is smaller than the preset area threshold or the distance threshold is greater than the preset distance threshold, the corresponding area of ​​interest is adjusted according to the area of ​​the region to obtain a corrected area of ​​interest for the corresponding area of ​​interest; if the area of ​​the region is larger than the preset area threshold and the distance threshold is smaller than the preset distance threshold, the corrected fault areas of the corresponding areas of interest are merged, and the merged corrected fault areas are set as the corrected area of ​​interest.

[0030] In some embodiments of the present application, a fault association model for each modified area of ​​interest is constructed according to the fault association relationship between the modified areas of interest, including:

[0031] Obtaining a fault time node in the fault characteristics of each corrected focus area, setting each corrected focus area as a first center, and subtracting the fault time node of the first center from the fault time nodes of the remaining corrected focus areas to obtain a fault time difference;

[0032] Presetting a first preset time difference interval and a second preset time difference interval, wherein the first preset time difference interval is smaller than the second preset time difference interval;

[0033] When the fault time difference is within the first preset time difference interval, setting the corresponding modified focus area to be the first associated subset of the corresponding first center;

[0034] When the fault time difference is within a second preset time difference interval, setting the corresponding modified focus area to be a second associated subset of the corresponding first center;

[0035] Obtaining the fault degree of the modified focus area in the first associated subset and the second associated subset, and eliminating the modified focus area whose fault degree is less than a preset fault degree threshold;

[0036] Obtain the first intersection fault period of the remaining modified focus area in the first associated subset and the corresponding first center, construct a second time reference line according to the first intersection fault period, set a second acquisition node according to a second preset time interval, obtain characteristic fault data of the remaining modified focus area in the first associated subset and characteristic fault data of the first center according to the second acquisition node, and map them to the second time reference line;

[0037] Calculate the first degree of dependence of the characteristic fault data of the first center at each second acquisition node on the second time reference line and the characteristic fault data of the remaining corrected focus area in the first associated subset, and generate a first correlation coefficient between the characteristic fault data of the first center and the characteristic fault data of the remaining corrected focus area in the first associated subset according to the number of second acquisition nodes whose first degree of dependence is greater than the first preset dependence degree threshold and the first dependence degree difference value whose first degree of dependence is greater than the first preset dependence degree threshold;

[0038] Setting the characteristic fault data of the corrected focus area remaining in the first correlation subset whose first correlation coefficient is greater than the first preset correlation coefficient threshold as the first associated fault data of the corresponding characteristic fault data of the first center;

[0039] Obtain the second intersection fault period of the remaining corrected focus area in the second associated subset and the corresponding first center, construct a third time reference line according to the second intersection fault period, set a third acquisition node according to a third preset time interval, obtain characteristic fault data of the remaining corrected focus area in the second associated subset and characteristic fault data of the first center according to the third acquisition node, and map them to the third time reference line;

[0040] Calculate the second degree of dependence between the characteristic fault data of the first center at each third acquisition node on the third time reference line and the characteristic fault data of the remaining corrected focus area in the second associated subset, and generate a second correlation coefficient between the characteristic fault data of the first center and the characteristic fault data of the remaining corrected focus area in the second associated subset according to the number of third acquisition nodes whose second degree of dependence is greater than the second preset dependence degree threshold and the second dependence degree difference value whose second degree of dependence is greater than the second preset dependence degree threshold;

[0041] Setting the characteristic fault data of the corrected focus area remaining in the second correlation subset with a second correlation coefficient greater than a second preset correlation coefficient threshold as the second associated fault data of the corresponding characteristic fault data of the first center;

[0042] A fault association model corresponding to the modified focus area of ​​each first center is constructed according to the first associated subset, the second associated subset, the first associated fault data and the second associated fault data of each first center.

[0043] In some embodiments of the present application, constructing a fault association model corresponding to each first center and correcting the focus area includes:

[0044] Using the characteristic fault data of each first center as a training input parameter set, using the remaining corrected regions of interest in the first associated subset and the second associated subset of each first center as a first training output parameter set, using the first associated fault data and the second associated fault data corresponding to the characteristic fault data of each first center as a second training output parameter set, and mapping the corrected regions of interest in the first training output parameter set outputted at the same time with the first associated fault data and the second associated fault data of the second training output parameter set;

[0045] Based on the training input parameter set, the first training output parameter set and the second training output parameter set, neural network training is performed to generate a fault association model of a modified area of ​​interest corresponding to each first center.

[0046] In some embodiments of the present application, a state evaluation is performed on a corresponding correction focus area based on real-time blade parameters and a fault association model, including:

[0047] Based on the source and collection method of characteristic fault data of each correction focus area, a preset monitoring node is set, and a corresponding monitoring device is set at the preset monitoring node;

[0048] Acquire multiple real-time blade data at the monitoring device of each corrected focus area, and perform similarity calculation between the real-time blade data and characteristic fault data of different fault types corresponding to the corrected focus area, and generate a first fault coefficient corresponding to the real-time blade data according to the fault degree and fault influence coefficient of the characteristic fault data with the greatest similarity;

[0049] Inputting the real-time blade data into the fault correlation model corresponding to the modified focus area to obtain a first real-time correlation subset, a second real-time correlation subset, first real-time correlation fault data, and second real-time correlation fault data of the real-time blade data;

[0050] Mapping the first real-time associated subset with the first real-time associated fault data, and mapping the second real-time associated subset with the second real-time associated fault data, determining predicted fault probabilities of the corrected focus areas in the first real-time associated subset and the second real-time associated subset according to the mapping results, and generating a second fault coefficient of the real-time blade data according to the predicted fault probability;

[0051] A status evaluation is performed on the corresponding correction focus area according to the first fault coefficient and the second fault coefficient.

[0052] In some embodiments of the present application, the calculation formula for the status evaluation is:

[0053]

[0054] Among them, K is the state evaluation result, r is the state evaluation conversion coefficient, m is the total number of real-time blade data, n1 is the number of characteristic fault data whose similarity between real-time blade data and characteristic fault data of different fault types of corresponding corrected attention areas is greater than the preset similarity threshold, ti is the fault degree of the i-th characteristic fault data, qi is the fault influence coefficient of the i-th characteristic fault data, u1 is the weight coefficient of the first fault coefficient, u2 is the weight coefficient of the second fault coefficient, n2 is the total number of corrected attention areas in the first real-time associated subset, ws is the predicted failure probability of the s-th corrected attention area in the first real-time associated subset, d1s is the weight coefficient of the s-th corrected attention area in the first real-time associated subset, n3 is the total number of corrected attention areas in the second real-time associated subset, yz is the predicted failure probability of the z-th corrected attention area in the second real-time associated subset, and d2z is the weight coefficient of the z-th corrected attention area in the second real-time associated subset.

[0055] In some embodiments of the present application, determining whether to generate an alarm signal according to the status evaluation result includes:

[0056] Pre-set status evaluation thresholds;

[0057] If the status evaluation result is less than the status evaluation threshold, no alarm signal is generated;

[0058] If the status evaluation result is greater than the status evaluation threshold, an alarm signal is generated, and the alarm signal includes the location node of the correction focus area where the fault occurs and the fault type.

[0059] Compared with the prior art, the method for analyzing the failure of the last stage blade of the low pressure cylinder of a steam turbine in the embodiment of the present application has the following beneficial effects:

[0060] By establishing a blade structure map to calibrate multiple areas of concern, the fault area and fault characteristics of each area of ​​concern under real-time operating conditions are obtained, the corrected area of ​​concern is obtained, and a fault association model of the corrected area of ​​concern is constructed. The state evaluation is performed according to the real-time blade data and the fault association model to achieve accurate judgment of the blade state. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a flow chart of a method for analyzing failure of the last stage blades of a low-pressure cylinder of a steam turbine in a preferred embodiment of the present application. DETAILED DESCRIPTION

[0062] The specific implementation methods of the present application are further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application but are not intended to limit the scope of the present application.

[0063] In the description of the present application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0064] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0065] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0066] like Figure 1 As shown, a method for analyzing a fault of a last stage blade of a low-pressure cylinder of a steam turbine according to a preferred embodiment of the present application includes:

[0067] Step S101: establishing a blade structure atlas, wherein a plurality of regions of interest are marked in the blade structure atlas;

[0068] Step S102: acquiring historical fault information of each concerned area based on the real-time operating condition of the steam turbine, analyzing the historical fault information, and determining the fault area and fault characteristics of the concerned area;

[0069] Step S103: adjusting the corresponding focus areas based on the fault areas and fault characteristics to obtain modified focus areas, and constructing a fault association model for each modified focus area according to the fault association relationship between the modified focus areas;

[0070] Step S104: acquiring real-time blade data of each modified focus area, performing status evaluation on the corresponding modified focus area based on the real-time blade data and the fault association model, and determining whether to generate an alarm signal according to the status evaluation result.

[0071] In some embodiments of the present application, establishing a blade structure atlas includes:

[0072] Determine the model and real-time operating conditions of the steam turbine, and obtain the actual geometric parameters of the last-stage blades of the low-pressure cylinder of the steam turbine;

[0073] Construct the initial structural model of the last stage blade of the low pressure cylinder of the steam turbine according to the actual geometric parameters;

[0074] Evenly segment the direction from the root of the blade to one end of the blade in the initial structural model to obtain multiple focus areas in the initial structural model;

[0075] A reference coordinate system is established based on the aerodynamic center of the last stage blade of the low-pressure cylinder of the steam turbine under real-time operating conditions, and the vector coordinates of the critical position of each area of ​​interest are calibrated in the reference coordinate system;

[0076] The vector coordinates of the critical positions of different focus areas of the initial structural model are scaled proportionally to generate a blade structure map of the last-stage blade of the low-pressure cylinder of the steam turbine, and the corresponding focus areas are calibrated in the blade structure map.

[0077] In this embodiment, by constructing an initial structural model of the last-stage blades of the low-pressure cylinder of the turbine and dividing it into multiple areas of interest, the structure of the last-stage blades of the low-pressure cylinder of the turbine under real-time operating conditions is displayed, and proportional scaling is performed to obtain a blade structure map, laying the foundation for subsequent analysis of the blade status of each area of ​​interest, improving the analysis accuracy of the last-stage blades of the low-pressure cylinder of the turbine and the accuracy of fault judgment, thereby ensuring safe and stable operation of the unit.

[0078] In some embodiments of the present application, historical fault information is analyzed to determine the fault area and fault characteristics of the concerned area, including:

[0079] Obtain historical fault logs of the steam turbine, and classify the historical fault logs according to historical operating conditions corresponding to the historical fault logs to obtain historical fault logs of multiple categories;

[0080] According to the similarity between the real-time operating condition and the historical operating condition, the historical fault logs of the corresponding category are screened out, and the screened historical fault logs are analyzed to obtain the historical fault information of each concerned area of ​​the real-time operating condition, wherein the historical fault information includes the fault type, fault point, fault time node, fault data, fault degree and fault impact coefficient;

[0081] Determine the standard fault duration of the corresponding fault type in the corresponding area of ​​interest according to the fault time nodes in different historical fault information of the same fault type in each area of ​​interest;

[0082] Determine the fault area of ​​the corresponding fault type in the corresponding focus area according to the fault points of the same fault type in each focus area within multiple standard fault durations;

[0083] Establish a first time reference line according to the standard fault duration of the same fault type in each area of ​​concern, set a first acquisition node based on a first preset time interval, obtain fault data and fault degree within the same standard fault duration according to each first acquisition node, and map them to the first time reference line;

[0084] According to the degree of synchronous change of fault data and fault degree, characteristic fault data of corresponding fault type of each concerned area is screened out;

[0085] The standard fault durations, characteristic fault data and fault impact coefficients of multiple fault types in each area of ​​interest are set as the fault characteristics of the corresponding area of ​​interest.

[0086] In this embodiment, the synchronous change degree refers to the similarity of the change trends of the fault data and the fault degree at the same first acquisition node.

[0087] In this embodiment, based on determining the fault area and fault characteristics of each focus area, the areas on the blade where faults often occur are obtained, and the areas where faults often occur are monitored intensively, thereby improving the analysis accuracy and efficiency of the blade.

[0088] In some embodiments of the present application, the corresponding region of interest is modified based on the fault region and the fault characteristics to obtain a modified region of interest, including:

[0089] Comprehensively analyze the fault characteristics of different fault types in the fault area and generate a comprehensive fault evaluation value for the corresponding fault area;

[0090] According to the relationship between the comprehensive fault evaluation value and the preset fault evaluation value interval, a correction coefficient corresponding to the fault area is set, and a corrected fault area is generated;

[0091] Determine the area ratio of the corrected fault area to the corresponding area of ​​interest based on the positional relationship between the corrected fault area and the corresponding area of ​​interest;

[0092] If the area ratio is less than the first preset area ratio, setting the corrected fault area as a corrected focus area of ​​the corresponding focus area;

[0093] If the area ratio is greater than the first preset area ratio, determine whether the corrected fault area is entirely within the corresponding focus area; if so, do not adjust the corresponding focus area; if not, obtain the area areas of other focus areas where the corrected fault area is located, and calculate the distance threshold between the area boundary points of the other focus areas where the corrected fault area is located and the boundary points of the corrected fault areas of other focus areas;

[0094] If the area of ​​the region is smaller than the preset area threshold or the distance threshold is greater than the preset distance threshold, the corresponding area of ​​interest is adjusted according to the area of ​​the region to obtain a corrected area of ​​interest for the corresponding area of ​​interest; if the area of ​​the region is larger than the preset area threshold and the distance threshold is smaller than the preset distance threshold, the corrected fault areas of the corresponding areas of interest are merged, and the merged corrected fault areas are set as the corrected area of ​​interest.

[0095] In this embodiment, multiple preset fault evaluation value intervals are pre-set. When the preset fault evaluation value interval in which the comprehensive fault evaluation value is located is larger, the selected correction coefficient is larger. When the preset fault evaluation value interval in which the comprehensive fault evaluation value is located is smaller, the selected correction coefficient is smaller. The value range of the correction coefficient is (0.75, 1.25), and the corrected fault area = correction coefficient * fault area.

[0096] In this embodiment, each focus area is adjusted according to the fault area and fault characteristics to obtain a corrected focus area, thereby determining the area in the blade that is prone to faults, improving the monitoring efficiency and analysis efficiency, accelerating the status analysis of the last stage blades of the low-pressure cylinder of the turbine, and ensuring the safe and stable operation of the unit.

[0097] In some embodiments of the present application, a fault association model for each modified area of ​​interest is constructed according to the fault association relationship between the modified areas of interest, including:

[0098] Obtaining a fault time node in the fault characteristics of each corrected focus area, setting each corrected focus area as a first center, and subtracting the fault time node of the first center from the fault time nodes of the remaining corrected focus areas to obtain a fault time difference;

[0099] Presetting a first preset time difference interval and a second preset time difference interval, wherein the first preset time difference interval is smaller than the second preset time difference interval;

[0100] When the fault time difference is within the first preset time difference interval, setting the corresponding modified focus area to be the first associated subset of the corresponding first center;

[0101] When the fault time difference is within a second preset time difference interval, setting the corresponding modified focus area to be a second associated subset of the corresponding first center;

[0102] Obtaining the fault degree of the modified focus area in the first associated subset and the second associated subset, and eliminating the modified focus area whose fault degree is less than a preset fault degree threshold;

[0103] Obtain the first intersection fault period of the remaining modified focus area in the first associated subset and the corresponding first center, construct a second time reference line according to the first intersection fault period, set a second acquisition node according to a second preset time interval, obtain characteristic fault data of the remaining modified focus area in the first associated subset and characteristic fault data of the first center according to the second acquisition node, and map them to the second time reference line;

[0104] Calculate the first degree of dependence of the characteristic fault data of the first center at each second acquisition node on the second time reference line and the characteristic fault data of the remaining corrected focus area in the first associated subset, and generate a first correlation coefficient between the characteristic fault data of the first center and the characteristic fault data of the remaining corrected focus area in the first associated subset according to the number of second acquisition nodes whose first degree of dependence is greater than the first preset dependence degree threshold and the first dependence degree difference value whose first degree of dependence is greater than the first preset dependence degree threshold;

[0105] Setting the characteristic fault data of the corrected focus area remaining in the first correlation subset whose first correlation coefficient is greater than the first preset correlation coefficient threshold as the first associated fault data of the corresponding characteristic fault data of the first center;

[0106] Obtain the second intersection fault period of the remaining corrected focus area in the second associated subset and the corresponding first center, construct a third time reference line according to the second intersection fault period, set a third acquisition node according to a third preset time interval, obtain characteristic fault data of the remaining corrected focus area in the second associated subset and characteristic fault data of the first center according to the third acquisition node, and map them to the third time reference line;

[0107] Calculate the second degree of dependence between the characteristic fault data of the first center at each third acquisition node on the third time reference line and the characteristic fault data of the remaining corrected focus area in the second associated subset, and generate a second correlation coefficient between the characteristic fault data of the first center and the characteristic fault data of the remaining corrected focus area in the second associated subset according to the number of third acquisition nodes whose second degree of dependence is greater than the second preset dependence degree threshold and the second dependence degree difference value whose second degree of dependence is greater than the second preset dependence degree threshold;

[0108] Setting the characteristic fault data of the corrected focus area remaining in the second correlation subset with a second correlation coefficient greater than a second preset correlation coefficient threshold as the second associated fault data of the corresponding characteristic fault data of the first center;

[0109] A fault association model corresponding to the modified focus area of ​​each first center is constructed according to the first associated subset, the second associated subset, the first associated fault data and the second associated fault data of each first center.

[0110] In this embodiment, the fault time node includes a fault start time node and a fault end time node, the fault time difference = the fault start time node of the first center - the fault start time node of the remaining correction attention area, the first preset time difference interval <0 < the second preset time difference interval, when the fault time difference is less than 0, that is, the fault start time node of the first center is between the fault start time nodes of the remaining correction attention area, that is, the correction attention area corresponding to the first center fails first, when the fault time difference is greater than 0, that is, the fault start time node of the first center is after the fault start time node of the remaining correction attention area, that is, the remaining correction attention area in the first associated subset fails first.

[0111] In this embodiment, the first associated subset refers to a set of correction focus areas where failure in the first center causes failure in other correction focus areas, and the second associated subset refers to a set of correction focus areas where failure in other correction focus areas causes failure in the first center.

[0112] In this embodiment, the first intersection fault period and the second intersection fault period refer to fault periods in which the first center overlaps with the modified focus areas in the first associated subset and the second associated subset, respectively.

[0113] In this embodiment, the first degree of dependence refers to the degree to which the characteristic fault data of the remaining corrected focus area in the first associated subset changes with the change of the characteristic fault data of the first center, and the second degree of dependence refers to the degree to which the characteristic fault data of the first center changes with the change of the characteristic fault data of the remaining corrected focus area in the second associated subset.

[0114] In this embodiment, by analyzing the fault association relationship between each corrected focus area and other corrected focus areas, the first associated subset, the second associated subset, the first associated fault data and the second associated fault data of each corrected focus area are obtained, thereby laying a data foundation for the subsequent construction of a fault association model, improving the analysis accuracy of the fault of the last stage blade of the low-pressure cylinder of the turbine, clearly understanding the association relationship when a fault occurs in the corrected focus area, improving the accuracy of fault analysis, and improving the safety of the unit.

[0115] In some embodiments of the present application, constructing a fault association model corresponding to each first center and correcting the focus area includes:

[0116] Using the characteristic fault data of each first center as a training input parameter set, using the remaining corrected regions of interest in the first associated subset and the second associated subset of each first center as a first training output parameter set, using the first associated fault data and the second associated fault data corresponding to the characteristic fault data of each first center as a second training output parameter set, and mapping the corrected regions of interest in the first training output parameter set outputted at the same time with the first associated fault data and the second associated fault data of the second training output parameter set;

[0117] Based on the training input parameter set, the first training output parameter set and the second training output parameter set, neural network training is performed to generate a fault association model of a modified area of ​​interest corresponding to each first center.

[0118] In some embodiments of the present application, a state evaluation is performed on a corresponding correction focus area based on real-time blade parameters and a fault association model, including:

[0119] Based on the source and collection method of characteristic fault data of each correction focus area, a preset monitoring node is set, and a corresponding monitoring device is set at the preset monitoring node;

[0120] Acquire multiple real-time blade data at the monitoring device of each corrected focus area, and perform similarity calculation between the real-time blade data and characteristic fault data of different fault types corresponding to the corrected focus area, and generate a first fault coefficient corresponding to the real-time blade data according to the fault degree and fault influence coefficient of the characteristic fault data with the greatest similarity;

[0121] Inputting the real-time blade data into the fault correlation model corresponding to the modified focus area to obtain a first real-time correlation subset, a second real-time correlation subset, first real-time correlation fault data, and second real-time correlation fault data of the real-time blade data;

[0122] Mapping the first real-time associated subset with the first real-time associated fault data, and mapping the second real-time associated subset with the second real-time associated fault data, determining predicted fault probabilities of the corrected focus areas in the first real-time associated subset and the second real-time associated subset according to the mapping results, and generating a second fault coefficient of the real-time blade data according to the predicted fault probability;

[0123] A status evaluation is performed on the corresponding correction focus area according to the first fault coefficient and the second fault coefficient.

[0124] In this embodiment, a mapping relationship between the corrected area of ​​interest in the first real-time associated fault data and the corrected area of ​​interest in the second real-time associated fault data is obtained according to the mapping results, and the corresponding fault type and fault degree are determined according to the first real-time associated fault data or the second real-time associated fault data of the corrected area of ​​interest, and a predicted fault probability is generated according to the fault type and fault degree.

[0125] In this embodiment, the first fault coefficient of each corrected focus area itself is determined according to the similarity between the real-time blade data and the characteristic fault data, and the second fault coefficient of the corrected focus area fault for other corrected focus areas is obtained according to the fault association model. The status of the last stage blade of the low-pressure cylinder of the turbine is comprehensively evaluated based on the first fault coefficient and the second fault coefficient to improve the accuracy of fault status analysis.

[0126] In some embodiments of the present application, the calculation formula for the status evaluation is:

[0127]

[0128] Among them, K is the state evaluation result, r is the state evaluation conversion coefficient, m is the total number of real-time blade data, n1 is the number of characteristic fault data whose similarity between real-time blade data and characteristic fault data of different fault types of corresponding corrected attention areas is greater than the preset similarity threshold, ti is the fault degree of the i-th characteristic fault data, qi is the fault influence coefficient of the i-th characteristic fault data, u1 is the weight coefficient of the first fault coefficient, u2 is the weight coefficient of the second fault coefficient, n2 is the total number of corrected attention areas in the first real-time associated subset, ws is the predicted failure probability of the s-th corrected attention area in the first real-time associated subset, d1s is the weight coefficient of the s-th corrected attention area in the first real-time associated subset, n3 is the total number of corrected attention areas in the second real-time associated subset, yz is the predicted failure probability of the z-th corrected attention area in the second real-time associated subset, and d2z is the weight coefficient of the z-th corrected attention area in the second real-time associated subset.

[0129] In this embodiment, a first fault coefficient is generated by the similarity between real-time blade data and characteristic fault data, the predicted fault probability of real-time blade data for other corrected focus areas is obtained according to the fault association model, and a second fault coefficient is obtained. The status evaluation result of each corrected focus area is obtained according to the first fault coefficient and the second fault coefficient. When the status evaluation result is larger, it indicates that the probability of failure in the corresponding corrected focus area is greater and the probability of failure in other corrected focus areas is greater.

[0130] In some embodiments of the present application, determining whether to generate an alarm signal according to the status evaluation result includes:

[0131] Pre-set status evaluation thresholds;

[0132] If the status evaluation result is less than the status evaluation threshold, no alarm signal is generated;

[0133] If the status evaluation result is greater than the status evaluation threshold, an alarm signal is generated, and the alarm signal includes the location node of the correction focus area where the fault occurs and the fault type.

[0134] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and substitutions can be made without departing from the technical principles of the present application. These improvements and substitutions should also be regarded as the scope of protection of the present application.

Claims

1. A method for analyzing the fault of the last stage blade of the low pressure cylinder of a steam turbine, characterized in that: include: Establishing a leaf structure atlas, wherein a plurality of regions of interest are marked in the leaf structure atlas; Based on the real-time operating conditions of the steam turbine, historical fault information of each area of ​​concern is obtained, the historical fault information is analyzed, and the fault area and fault characteristics of the area of ​​concern are determined; Adjust the corresponding focus area based on the fault area and fault characteristics to obtain the modified focus area, and build a fault association model for each modified focus area according to the fault association relationship between the modified focus areas; The real-time blade data of each modified focus area is obtained, and the status of the corresponding modified focus area is evaluated based on the real-time blade data and the fault association model, and it is determined whether to generate an alarm signal according to the status evaluation result.

2. The method for analyzing the failure of the last blade of the low-pressure cylinder of a steam turbine according to claim 1, characterized in that: Establish a leaf structure map, including: Determine the model and real-time operating conditions of the steam turbine, and obtain the actual geometric parameters of the last-stage blades of the low-pressure cylinder of the steam turbine; Construct the initial structural model of the last stage blade of the low pressure cylinder of the steam turbine according to the actual geometric parameters; Evenly segment the direction from the root of the blade to one end of the blade in the initial structural model to obtain multiple focus areas in the initial structural model; A reference coordinate system is established based on the aerodynamic center of the last stage blade of the low-pressure cylinder of the steam turbine under real-time operating conditions, and the vector coordinates of the critical position of each area of ​​interest are calibrated in the reference coordinate system; The vector coordinates of the critical positions of different focus areas of the initial structural model are scaled proportionally to generate a blade structure map of the last-stage blade of the low-pressure cylinder of the steam turbine, and the corresponding focus areas are calibrated in the blade structure map.

3. The method for analyzing the failure of the last blade of the low-pressure cylinder of a steam turbine according to claim 2, characterized in that: Analyze historical fault information to determine the fault area and fault characteristics of the concerned area, including: Obtain historical fault logs of the steam turbine, and classify the historical fault logs according to historical operating conditions corresponding to the historical fault logs to obtain historical fault logs of multiple categories; According to the similarity between the real-time operating condition and the historical operating condition, the historical fault logs of the corresponding category are screened out, and the screened historical fault logs are analyzed to obtain the historical fault information of each concerned area of ​​the real-time operating condition, wherein the historical fault information includes the fault type, fault point, fault time node, fault data, fault degree and fault impact coefficient; Determine the standard fault duration of the corresponding fault type in the corresponding area of ​​interest according to the fault time nodes in different historical fault information of the same fault type in each area of ​​interest; Determine the fault area of ​​the corresponding fault type in the corresponding focus area according to the fault points of the same fault type in each focus area within multiple standard fault durations; Establish a first time reference line according to the standard fault duration of the same fault type in each area of ​​concern, set a first acquisition node based on a first preset time interval, obtain fault data and fault degree within the same standard fault duration according to each first acquisition node, and map them to the first time reference line; According to the degree of synchronous change of fault data and fault degree, characteristic fault data of corresponding fault type of each concerned area is screened out; The standard fault durations, characteristic fault data and fault impact coefficients of multiple fault types in each area of ​​interest are set as the fault characteristics of the corresponding area of ​​interest.

4. The method for analyzing the failure of the last stage blade of the low pressure cylinder of a steam turbine according to claim 3, characterized in that: Based on the fault area and fault characteristics, the corresponding focus area is corrected to obtain a corrected focus area, including: Comprehensively analyze the fault characteristics of different fault types in the fault area and generate a comprehensive fault evaluation value for the corresponding fault area; According to the relationship between the comprehensive fault evaluation value and the preset fault evaluation value interval, a correction coefficient corresponding to the fault area is set, and a corrected fault area is generated; Determine the area ratio of the corrected fault area to the corresponding area of ​​interest based on the positional relationship between the corrected fault area and the corresponding area of ​​interest; If the area ratio is less than the first preset area ratio, setting the corrected fault area as a corrected focus area of ​​the corresponding focus area; If the area ratio is greater than the first preset area ratio, determine whether the corrected fault area is entirely within the corresponding focus area; if so, do not adjust the corresponding focus area; if not, obtain the area areas of other focus areas where the corrected fault area is located, and calculate the distance threshold between the area boundary points of the other focus areas where the corrected fault area is located and the boundary points of the corrected fault areas of other focus areas; If the area of ​​the region is smaller than the preset area threshold or the distance threshold is greater than the preset distance threshold, the corresponding area of ​​interest is adjusted according to the area of ​​the region to obtain a corrected area of ​​interest for the corresponding area of ​​interest; if the area of ​​the region is larger than the preset area threshold and the distance threshold is smaller than the preset distance threshold, the corrected fault areas of the corresponding areas of interest are merged, and the merged corrected fault areas are set as the corrected area of ​​interest.

5. The method for analyzing the failure of the last stage blade of the low pressure cylinder of a steam turbine according to claim 4, characterized in that: According to the fault correlation relationship between the revised focus areas, a fault correlation model for each revised focus area is constructed, including: Obtaining a fault time node in the fault characteristics of each corrected focus area, setting each corrected focus area as a first center, and subtracting the fault time node of the first center from the fault time nodes of the remaining corrected focus areas to obtain a fault time difference; Presetting a first preset time difference interval and a second preset time difference interval, wherein the first preset time difference interval is smaller than the second preset time difference interval; When the fault time difference is within the first preset time difference interval, setting the corresponding modified focus area to be the first associated subset of the corresponding first center; When the fault time difference is within a second preset time difference interval, setting the corresponding modified focus area to be a second associated subset of the corresponding first center; Obtaining the fault degree of the modified focus area in the first associated subset and the second associated subset, and eliminating the modified focus area whose fault degree is less than a preset fault degree threshold; Obtain the first intersection fault period of the remaining modified focus area in the first associated subset and the corresponding first center, construct a second time reference line according to the first intersection fault period, set a second acquisition node according to a second preset time interval, obtain characteristic fault data of the remaining modified focus area in the first associated subset and characteristic fault data of the first center according to the second acquisition node, and map them to the second time reference line; Calculate the first degree of dependence of the characteristic fault data of the first center at each second acquisition node on the second time reference line and the characteristic fault data of the remaining corrected focus area in the first associated subset, and generate a first correlation coefficient between the characteristic fault data of the first center and the characteristic fault data of the remaining corrected focus area in the first associated subset according to the number of second acquisition nodes whose first degree of dependence is greater than the first preset dependence degree threshold and the first dependence degree difference value whose first degree of dependence is greater than the first preset dependence degree threshold; Setting the characteristic fault data of the corrected focus area remaining in the first correlation subset whose first correlation coefficient is greater than the first preset correlation coefficient threshold as the first associated fault data of the corresponding characteristic fault data of the first center; Obtain the second intersection fault period of the remaining corrected focus area in the second associated subset and the corresponding first center, construct a third time reference line according to the second intersection fault period, set a third acquisition node according to a third preset time interval, obtain characteristic fault data of the remaining corrected focus area in the second associated subset and characteristic fault data of the first center according to the third acquisition node, and map them to the third time reference line; Calculate the second degree of dependence between the characteristic fault data of the first center at each third acquisition node on the third time reference line and the characteristic fault data of the remaining corrected focus area in the second associated subset, and generate a second correlation coefficient between the characteristic fault data of the first center and the characteristic fault data of the remaining corrected focus area in the second associated subset according to the number of third acquisition nodes whose second degree of dependence is greater than the second preset dependence degree threshold and the second dependence degree difference value whose second degree of dependence is greater than the second preset dependence degree threshold; Setting the characteristic fault data of the corrected focus area remaining in the second correlation subset with a second correlation coefficient greater than a second preset correlation coefficient threshold as the second associated fault data of the corresponding characteristic fault data of the first center; A fault association model corresponding to the modified focus area of ​​each first center is constructed according to the first associated subset, the second associated subset, the first associated fault data and the second associated fault data of each first center.

6. The method for analyzing the failure of the last stage blade of the low pressure cylinder of a steam turbine according to claim 5, characterized in that: Construct a fault correlation model for each first center corresponding to the correction focus area, including: Using the characteristic fault data of each first center as a training input parameter set, using the remaining corrected regions of interest in the first associated subset and the second associated subset of each first center as a first training output parameter set, using the first associated fault data and the second associated fault data corresponding to the characteristic fault data of each first center as a second training output parameter set, and mapping the corrected regions of interest in the first training output parameter set outputted at the same time with the first associated fault data and the second associated fault data of the second training output parameter set; Based on the training input parameter set, the first training output parameter set and the second training output parameter set, neural network training is performed to generate a fault association model of a modified area of ​​interest corresponding to each first center.

7. The method for analyzing the failure of the last stage blade of the low pressure cylinder of a steam turbine according to claim 6, characterized in that: Based on real-time blade parameters and fault association models, the status of the corresponding correction focus areas is evaluated, including: Based on the source and collection method of characteristic fault data of each correction focus area, a preset monitoring node is set, and a corresponding monitoring device is set at the preset monitoring node; Acquire multiple real-time blade data at the monitoring device of each corrected focus area, and perform similarity calculation between the real-time blade data and characteristic fault data of different fault types corresponding to the corrected focus area, and generate a first fault coefficient corresponding to the real-time blade data according to the fault degree and fault influence coefficient of the characteristic fault data with the greatest similarity; Inputting the real-time blade data into the fault correlation model corresponding to the modified focus area to obtain a first real-time correlation subset, a second real-time correlation subset, first real-time correlation fault data, and second real-time correlation fault data of the real-time blade data; Mapping the first real-time associated subset with the first real-time associated fault data, and mapping the second real-time associated subset with the second real-time associated fault data, determining predicted fault probabilities of the corrected focus areas in the first real-time associated subset and the second real-time associated subset according to the mapping results, and generating a second fault coefficient of the real-time blade data according to the predicted fault probability; A status evaluation is performed on the corresponding correction focus area according to the first fault coefficient and the second fault coefficient.

8. The method for analyzing the fault of the last blade of the low-pressure cylinder of a steam turbine according to claim 7, characterized in that: The calculation formula for the status evaluation is: Among them, K is the state evaluation result, r is the state evaluation conversion coefficient, m is the total number of real-time blade data, n1 is the number of characteristic fault data whose similarity between real-time blade data and characteristic fault data of different fault types of corresponding corrected attention areas is greater than the preset similarity threshold, ti is the fault degree of the i-th characteristic fault data, qi is the fault influence coefficient of the i-th characteristic fault data, u1 is the weight coefficient of the first fault coefficient, u2 is the weight coefficient of the second fault coefficient, n2 is the total number of corrected attention areas in the first real-time associated subset, ws is the predicted failure probability of the s-th corrected attention area in the first real-time associated subset, d1s is the weight coefficient of the s-th corrected attention area in the first real-time associated subset, n3 is the total number of corrected attention areas in the second real-time associated subset, yz is the predicted failure probability of the z-th corrected attention area in the second real-time associated subset, and d2z is the weight coefficient of the z-th corrected attention area in the second real-time associated subset.

9. The method for analyzing the fault of the last blade of the low-pressure cylinder of a steam turbine according to claim 8, characterized in that: Determine whether to generate an alarm signal based on the status evaluation results, including: Pre-set status evaluation thresholds; If the status evaluation result is less than the status evaluation threshold, no alarm signal is generated; If the status evaluation result is greater than the status evaluation threshold, an alarm signal is generated, and the alarm signal includes the location node of the correction focus area where the fault occurs and the fault type.

Images