CN-121980245-A - Online identification method for vibration abnormality of steam turbine

CN121980245ACN 121980245 ACN121980245 ACN 121980245ACN-121980245-A

Abstract

The invention relates to the technical field of turbine vibration test and fault diagnosis and discloses a turbine vibration abnormality online identification method which comprises the steps of collecting vibration signals and key phase signals of all channels, calculating basic characteristics of all channels in a sliding window, establishing a redundant consistency graph, calculating consistency residual vectors, calculating graph consistency cost for each node, calculating comprehensive unreliable scores and reliability weights, constructing reference characteristic vectors, calculating pseudo vibration residual for each channel, carrying out pseudo vibration priority judgment, calculating pseudo vibration scores, recalculating reference characteristic vectors and calculating mechanical abnormality scores after eliminating the pseudo vibration priority channels, carrying out abnormality judgment and outputting judgment results.

Inventors

CHEN XIJIANG
FU QIANG
WEI ZHENG
WANG WEI
LIU LI

Assignees

国能常州发电有限公司
国能常州第二发电有限公司

Dates

Publication Date: 20260505
Application Date: 20260403

Claims (10)

1. The online identification method for the vibration abnormality of the steam turbine is characterized by comprising the following steps of: Collecting vibration signals and key phase signals of all channels, evaluating and processing the quality of the key phase signals, and calculating basic characteristics of all channels in a sliding window; Establishing a redundant consistency graph, calculating a consistency residual vector based on the redundant consistency graph and basic characteristics, calculating a graph consistency cost for each node, and calculating a comprehensive unreliable score and a reliability weight based on the graph consistency cost; weighting and steady fusion are carried out on the basic features of a plurality of channels to construct reference feature vectors, a pseudo vibration residual error is calculated for each channel based on the reference feature vectors and the basic features, and the pseudo vibration priority judgment is carried out; Calculating a pseudo vibration score based on the comprehensive unreliable score and the pseudo vibration residual error, and recalculating a reference feature vector and calculating a mechanical anomaly score after eliminating a pseudo vibration priority channel; and performing abnormality determination based on the pseudo vibration score and the mechanical abnormality score and outputting a determination result.
2. The method for online identification of turbine vibration anomalies according to claim 1, wherein evaluating and processing key phase signal quality comprises: Performing time alignment processing on the collected vibration signals of all channels and key phase signals; evaluating the key phase signal quality includes calculating a coefficient of variation of adjacent key phase pulse time intervals, and judging that the key phase signal does not pass when the coefficient of variation exceeds a preset key phase quality threshold value, or else, the key phase signal passes; when the key phase signal quality evaluation is not passed, maintaining a time domain analysis mode, and extracting a frequency conversion related characteristic by adopting short-time Fourier transform; the basic features include overall amplitude, order amplitude, band energy ratio, kurtosis, and spectral kurtosis.
3. The method of claim 1, wherein calculating a reliability weight based on the map consistency cost comprises: Taking each channel as a node, establishing a redundant consistency graph, wherein the redundant consistency graph comprises a node set and an edge set, the edge set comprises a plurality of edges, a connecting edge is established between two channels in the orthogonal direction of a coaxial bearing, a connecting edge is established between channels of the same type of sensors on adjacent bearings, and if a cross-type channel of shaft vibration and tile vibration or shell vibration exists, the connecting edge is established; calculating a consistency residual vector for each edge in the redundant consistency graph, wherein the consistency residual vector comprises an order amplitude consistency residual, a coherence consistency residual and a phase difference stability residual; carrying out standardization processing on the consistency residual error vector; calculating a graph consistency cost for each node based on the redundant consistency graph; calculating a composite unreliable score based on the graph consistency cost; the composite unreliable score is mapped to a reliability weight.
4. The online identification method of turbine vibration anomalies according to claim 3, wherein the order amplitude consistency residual is: Judging whether frequency multiplication amplitudes of two connected channels are both larger than a preset amplitude threshold or not; when the frequency multiplication amplitude values of the two nodes are larger than the preset amplitude threshold, respectively adding a preset small normal number constant to the frequency multiplication amplitude values of the two nodes to obtain sum values, respectively taking natural logarithms to the two sum values, and calculating the absolute value of the difference between the two logarithm values as an order amplitude consistency residual; when the frequency multiplication amplitude of any node is lower than the preset amplitude threshold, dividing the absolute value of the difference between the frequency multiplication amplitudes of the two nodes by an amplitude term, wherein the amplitude term is the sum of the frequency multiplication amplitudes of the two nodes plus a preset small normal number constant, and the sum is used as an order amplitude consistency residual.
5. The online identification method of the turbine vibration anomalies according to claim 3, wherein the coherent consistency residual is obtained by calculating coherent coefficients of two connected nodes at a plurality of characteristic frequencies, wherein the characteristic frequencies comprise frequency doubling, frequency doubling and frequency half; The phase difference stability residual error is obtained by respectively obtaining phase value sequences of two connected channels at a frequency multiplication position, calculating phase difference sequences of the two phase value sequences, performing phase expansion processing on the phase difference sequences to eliminate positive and negative circumferential rate boundary jump, and calculating the variance of the expanded phase difference sequences to be used as the phase difference stability residual error.
6. A method of online identification of turbine vibration anomalies according to claim 3, wherein calculating a graph consistency cost for each node comprises: Traversing all edges connected with the node, summing absolute values of components of the consistency residual vector for each edge to obtain a residual norm of the edge, multiplying the residual norm by preset edge weight to obtain a weighted residual norm, and accumulating the weighted residual norms of all the connected edges to obtain the graph consistency cost of the node.
7. The method of claim 1, wherein calculating a pseudo-vibration residual for each channel based on the reference feature vector and the base feature, and performing a pseudo-vibration priority determination comprises: In the same bearing or the same rotor section, weighting and steady fusion are carried out on basic features of a plurality of channels, and reference feature vectors are constructed, wherein the reference feature vectors are calculated by adopting a steady aggregation method based on Huber loss; calculating a pseudo vibration residual error for each channel, subtracting corresponding elements of the basic feature and the reference feature vector of the channel to obtain a difference vector, and calculating the Euclidean norm of the difference vector; and when the pseudo vibration residual error of the channel exceeds a preset residual error threshold and the reliability weight is lower than a preset weight threshold, judging that the abnormality of the channel is the pseudo vibration priority.
8. The method of claim 1, wherein calculating a pseudo vibration score comprises: carrying out weighted summation on the comprehensive unreliable score and the pseudo vibration residual error of each channel, and carrying out nonlinear transformation on the weighted summation to obtain a pseudo vibration tendency value of the channel; for the channel marked as the false vibration priority channel, the weighted sum is amplified when the false vibration tendency value is calculated; And taking the value of the preset quantile as a pseudo vibration score in the pseudo vibration tendency values of all the channels.
9. The method for online identification of turbine vibration anomalies according to claim 1, wherein calculating a mechanical anomaly score comprises: Removing the channel marked as the pseudo vibration priority channel, and recalculating the reference feature vector by using the rest channels; calculating trend increment of the reference feature vector, wherein the trend increment is obtained by calculating the difference value between the current reference feature vector and the historical baseline reference feature vector; And splicing the reference feature vector, the trend increment and the operation parameter basic feature into a vector, calculating an inner product of the vector and a preset weight vector, and carrying out nonlinear transformation on the inner product to obtain a mechanical anomaly score, wherein the operation parameter basic feature comprises a load, a main steam pressure, a main steam temperature, a current value of a back pressure operation parameter and a change rate of the current value.
10. The online identification method of turbine vibration anomalies according to claim 1, wherein making anomaly decisions based on the pseudo-vibration scores and mechanical anomaly scores and outputting the decision results comprises: outputting a judging result to be the abnormal or pseudo vibration of the measuring link when the pseudo vibration score is larger than or equal to a preset pseudo vibration judging threshold value and the mechanical abnormality score is smaller than the preset mechanical abnormality judging threshold value; Outputting a judging result as mechanical abnormal vibration when the mechanical abnormal score is larger than or equal to a preset mechanical abnormal judging threshold value and the pseudo vibration score is smaller than a preset pseudo vibration judging threshold value; when the pseudo vibration score is greater than or equal to a preset pseudo vibration discrimination threshold and the mechanical abnormality score is greater than or equal to a preset mechanical abnormality discrimination threshold, outputting a determination result as a suspected mechanical abnormality vibration accompanying measurement abnormality.

Description

Online identification method for vibration abnormality of steam turbine Technical Field The invention relates to the technical field of turbine vibration test and fault diagnosis, in particular to an online identification method for turbine vibration abnormality. Background The turbine unit is core rotating equipment of an electric power and industrial driving system, a rotor, a bearing and a basic system of the turbine unit continuously run under the working conditions of high rotating speed, high temperature and high pressure for a long time, and the turbine unit is easy to have the abnormal vibration problems of unbalance, poor centering, unstable oil film, rub-impact, loosening of parts and the like, and can cause unplanned shutdown when serious, so that great production loss is caused. In order to realize real-time monitoring and fault early warning of vibration states, sensors such as shaft vibration, tile vibration, shell vibration and the like are generally arranged at key positions of a bearing seat and a shaft system on site, an on-line monitoring and alarming system is built in a matched manner, vibration signals are acquired through devices such as an eddy current displacement sensor, a speed/acceleration sensor and the like, and trend analysis and anomaly judgment are carried out. The prior Chinese patent with the publication number of CN112557039B discloses a method for diagnosing abnormal vibration faults of a steam turbine with operation parameters coupled with vibration. The method comprises the following steps of S1, obtaining vibration characteristic information, S2, judging whether main vibration frequency of abnormal vibration is fundamental frequency, S3, judging the magnitude of a frequency multiplication value or a low-frequency vibration value in abnormal vibration, judging whether the ratio of the frequency multiplication value or the low-frequency vibration value in abnormal vibration to a pass frequency vibration value is larger than or equal to a set value, S4, judging whether the main vibration frequency of vibration is 2 frequency multiplication, S5, judging the magnitude of a frequency vibration peak value between the fundamental frequency and the 2 frequency multiplication value in abnormal vibration, S6, judging the magnitude of the vibration value in certain rotating speed, and S7, judging whether shaft vibration is far larger than tile vibration. However, in practical applications, due to complex working conditions and severe environmental effects, the measurement link is very prone to pseudo-vibration phenomena, i.e. the monitored vibration anomalies are not derived from the actual mechanical state of the rotor or the bearing, but are caused by faults of the sensor body, the mounting structure or the signal link. The expression form of the pseudo vibration is highly similar to that of a real mechanical fault, and the characteristics of amplitude rising, specific frequency band peak value enhancement, unstable phase and the like can occur, so that the prior art is easy to misjudge, and the pseudo vibration and the real mechanical fault cannot be effectively distinguished. Disclosure of Invention The invention aims to solve the problems and provide an online identification method for the vibration abnormality of a steam turbine. The invention provides a method for identifying the vibration abnormality of a steam turbine on line, which comprises the following steps: Collecting vibration signals and key phase signals of all channels, evaluating and processing the quality of the key phase signals, and calculating basic characteristics of all channels in a sliding window; Establishing a redundant consistency graph, calculating a consistency residual vector based on the redundant consistency graph and basic characteristics, calculating a graph consistency cost for each node, and calculating a comprehensive unreliable score and a reliability weight based on the graph consistency cost; weighting and steady fusion are carried out on the basic features of a plurality of channels to construct reference feature vectors, a pseudo vibration residual error is calculated for each channel based on the reference feature vectors and the basic features, and the pseudo vibration priority judgment is carried out; Calculating a pseudo vibration score based on the comprehensive unreliable score and the pseudo vibration residual error, and recalculating a reference feature vector and calculating a mechanical anomaly score after eliminating a pseudo vibration priority channel; and performing abnormality determination based on the pseudo vibration score and the mechanical abnormality score and outputting a determination result. Further, evaluating and processing the key phase signal quality includes: Performing time alignment processing on the collected vibration signals of all channels and key phase signals; evaluating the key phase signal quality includes calculating a coefficient of variation of