Method for evaluating vibration consistency of combustion chamber of gas turbine based on multi-modal fusion

Abstract

The invention relates to the technical field of gas turbine vibration evaluation, and discloses a gas turbine combustion chamber vibration consistency evaluation method based on multi-modal fusion, and the method comprises the steps: obtaining a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; carrying out segmentation processing on the data set to form sub-data sets; constructing a multi-modal feature vector based on the sub-data set; constructing a depth feature extraction model, and extracting deep features of the multi-modal feature vectors; clustering analysis is carried out on the deep features, health state grades of combustion chamber vibration are divided, and multi-grade health state grade criteria are formed; according to the real-time state signals of all the combustion chambers in the current operation state, the health state grade of current combustion chamber vibration is judged; according to the method, through multi-modal data fusion and depth feature extraction, the vibration consistency of the combustion chamber of the gas turbine can be evaluated more accurately, and the operation safety and reliability of the gas turbine are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas turbine vibration assessment, and more specifically, to a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion. Background Art

[0002] Gas turbines are important power equipment and are widely used in the fields of electricity, ships, etc. The vibration health of gas turbines is crucial to their safe and stable operation. As the core component of gas turbines, the vibration consistency of the combustion chamber has a key impact on the vibration health of the whole machine. At present, there has been some progress in the research on the vibration consistency evaluation of gas turbine combustion chambers based on multi-modal fusion, but there are still some shortcomings.

[0003] The Chinese patent with the authorization announcement number CN113688531B proposes a gas turbine reliability assessment method with two-dimensional vibration parameters containing measurement errors. This method takes into account the impact of sensor measurement errors on vibration test data processing, and uses the Frank Copula function to correlate two vibration test data to evaluate the reliability of the gas turbine core rotor system. However, this method is mainly aimed at the rotor system, and there is little research on the vibration consistency assessment of the combustion chamber. In addition, this method is mainly based on vibration signals, lacks consideration of other modal signals such as temperature, and the comprehensiveness of the assessment needs to be improved.

[0004] A Chinese patent with the authorization announcement number CN113176094B discloses an online monitoring system for thermoacoustic oscillations in gas turbine combustion chambers. The system collects multi-attribute status data such as pressure pulsation, vibration acceleration, and infrared high temperature in real time by installing multiple sensors on each flame tube, and performs multi-domain analysis and fault warning. The system can improve the timeliness and reliability of thermoacoustic oscillation monitoring and warning in the combustion chamber. However, the system mainly focuses on the phenomenon of thermoacoustic oscillation and has limited ability to evaluate general vibration problems. In addition, the system evaluates each combustion chamber independently and lacks an overall evaluation of vibration consistency.

[0005] In summary, the existing technology lacks a method to integrate multimodal information such as vibration and temperature in the gas turbine combustion chamber vibration consistency assessment based on multimodal fusion, and fails to evaluate the vibration consistency of multiple combustion chambers from a holistic perspective. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, the present invention provides a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion. The method obtains vibration and temperature signals of multiple combustion chambers, segments and extracts features from historical data, constructs a deep learning model to mine deep features, forms multi-level health status criteria, and finally realizes accurate assessment of the current combustion chamber vibration consistency; the method can fuse multi-modal information, deeply mine complex operating condition data, and evaluate the vibration consistency of multiple combustion chambers from an overall perspective.

[0007] To achieve the above object, the present invention provides the following technical solutions: The gas turbine combustor vibration consistency assessment method based on multi-modal fusion includes: A multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber is obtained; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; the historical combustion vibration data set is segmented to form n1 sub-data sets; for each sub-data set, a preliminary feature vector based on Mahalanobis distance is constructed, and based on the preliminary feature vector, a multi-modal feature vector based on mutual information weighted fusion is constructed; Construct a deep feature extraction model to extract deep features of multimodal feature vectors; perform cluster analysis on deep features to classify the health status level of combustion chamber vibration; calculate the cluster center of each health status level to form a multi-level health status level criterion; The real-time status signal of each combustion chamber under the current operating state is collected, and the health status level of the current combustion chamber vibration is determined according to the real-time status signal and the multi-level health status level criterion.

[0008] Furthermore, the segmentation processing of the historical combustion vibration data set to form n1 sub-data sets includes: Extracting gas turbine operating parameters corresponding to the historical combustion vibration data set, the operating parameters including load and speed; The load change threshold δ is set, and the historical combustion vibration data set is traversed and analyzed in time series. With time t0 as the starting point, if the load change at time t1 relative to the starting time t0 exceeds the load change threshold δ, it is considered that the operating condition has switched, and time t1 is marked as the operating condition switching point; then, with time t1 as the new starting point, the marking process of the operating condition switching point is repeated until the end of the historical combustion vibration data set is traversed; According to the operating condition switching point, the historical combustion vibration data set is divided into n2 initial operating condition segments; Cluster the operating parameters in the initial operating condition segment and divide the historical combustion vibration data set into n1 subdivided operating condition segments; A corresponding sub-dataset is generated for each subdivided operating condition segment, and n1 sub-datasets are obtained.

[0009] Furthermore, clustering the operating condition parameters in the initial operating condition section includes: The operating parameters in the initial operating condition section are used as sample features to form a feature matrix X. The dimension of the feature matrix X is m×n, where m is the number of samples and n is the number of features. Set the value range of cluster number K to ,in Indicates rounding down; traverse the cluster number K value and calculate the silhouette coefficient sequence and DBI index sequence; The comprehensive metric index sequence is calculated according to the silhouette coefficient sequence and the DBI index sequence, and the cluster number corresponding to the maximum value in the comprehensive metric index sequence is taken as the optimal cluster number N1; The optimal clustering number N1 is fixed, and the K-Means algorithm is used to cluster the feature matrix X to obtain n1 subdivided operating condition segments and the cluster center of each subdivided operating condition segment, forming the cluster center vector of the subdivided operating condition segment; where N1=n1.

[0010] Furthermore, the construction of a preliminary feature vector based on Mahalanobis distance includes: Calculate the Mahalanobis distance of the vibration signal of each combustion chamber relative to the mean of the sub-data set to form a vibration Mahalanobis distance vector ,in, represents the vibration Mahalanobis distance vector of combustion chamber j in the i-th operating condition, i represents the operating condition number, and j represents the combustion chamber number; Calculate the Mahalanobis distance of the temperature signal of each combustion chamber relative to the mean of the sub-dataset to form the temperature Mahalanobis distance vector ;in, represents the temperature Mahalanobis distance vector of combustion chamber j in the i-th operating condition; The vibration Mahalanobis distance vector and the temperature Mahalanobis distance vector are combined into a preliminary feature vector according to the combustion chamber sequence. , , Represents the combustion chamber of the i-th operating condition The initial feature vector of .

[0011] Furthermore, the construction of a multimodal feature vector based on mutual information weighted fusion includes: For each combustion chamber, the mutual information matrix of the vibration signal and the temperature signal is calculated ; represents the mutual information matrix between the vibration signal and the temperature signal in the i-th operating condition; Mutual Information Matrix Perform normalization to obtain the fusion weight matrix ; represents the fusion weight matrix of the i-th operating condition; According to the fusion weight matrix Weighted fusion of the i-th operating condition segment The vibration Mahalanobis distance vector of the combustion chamber and temperature Mahalanobis distance vector , construct multimodal fusion features ; Indicates the i-th operating condition segment Multi-modal fusion characteristics of a combustion chamber; The multimodal fusion features of each combustion chamber Composition Multi-mode feature vector of each operating condition .

[0012] Furthermore, the deep feature extraction model includes a convolutional neural network and an autoencoder; The deep features of the multimodal feature vector extraction include: The multimodal feature vector is input into the convolutional neural network to extract the high-dimensional feature map; the high-dimensional feature map output by the convolutional neural network is input into the autoencoder to learn the globally compressed low-dimensional features; the globally compressed low-dimensional features output by the autoencoder are used as the deep features of the multimodal feature vector.

[0013] Further, the health status level of the combustion chamber vibration includes normal, warning and abnormal; The calculation of the cluster center of each health status level to form a multi-level health status level criterion includes: Calculate the coordinates Cn, Cw, Ca of the cluster centers of the three health status levels of normal, warning, and abnormal respectively; Calculate the distance from each sample in the normal state cluster to the normal state cluster center coordinate Cn, and record the maximum distance as d1; Calculate the distance from each sample in the warning state cluster to the center coordinate Cn of the normal state cluster, and record the minimum distance as d2; let δ1=1 / 2(d1+d2) as the boundary threshold of the normal state; Calculate the distance from each sample in the warning state cluster to the coordinate Cw of the center of the warning state cluster, and record the maximum distance as d3; Calculate the distance from each sample in the abnormal state cluster to the center coordinate Cw of the warning state cluster, and record the minimum distance as d4; let δ2=1 / 2(d3+d4) as the boundary threshold of the warning state; Calculate the distance from each sample in the abnormal state cluster to the abnormal state cluster center coordinate Ca, record the maximum distance as d5, and set δ3=2×d5 as the boundary threshold of the abnormal state.

[0014] Further, the real-time status signal includes a real-time vibration signal and a real-time temperature signal; Determining the health status level of the current combustor vibration according to the real-time status signal and the multi-level health status level criterion includes: Determine the operating condition section to which the current operating status belongs, and mark it as the current operating condition section; calculate the real-time multi-modal feature vector of the real-time status signal; Input the real-time multi-modal feature vector into the deep feature extraction model to obtain the real-time deep feature; calculate the Euclidean distances Dn, Dw, and Da between the real-time deep feature and the coordinates Cn, Cw, and Ca of the clustering centers of the three health status levels of normal, warning, and abnormal, and determine the health status level of the current combustor vibration.

[0015] Further, the determining the operating condition section to which the current operating status belongs includes: Extract the real-time operating condition parameters under the current operating status; perform dimensionless processing on the real-time operating condition parameters to convert the real-time operating condition parameters into the real-time operating condition vector x (c) ; calculate the Euclidean distances between the real-time operating condition vector and the clustering center vectors of each clustering center in the subdivided operating condition sections; the real-time operating condition vector x (c) is divided into the subdivided operating condition section corresponding to the clustering center with the closest distance as the operating condition section of the current operating status; The calculating the real-time multi-modal feature vector of the real-time status signal includes: Calculate the Mahalanobis distance between the real-time vibration signal and the mean of the sub-dataset of the belonging operating condition section to form the real-time preliminary feature vector; according to the fusion weight matrix and the real-time preliminary feature vector, construct the real-time multi-modal feature vector based on mutual information weighted fusion.

[0016] Further, the determining the health status level of the current combustor vibration includes: If Dn≤δ1 and Dw≤Da, it is determined to be in the normal state; If δ1<Dn≤δ2 and Dw≤Da, it is determined to be in the warning state; If δ2<Dn≤δ3, it is determined to be in the abnormal state.

[0017] A gas turbine combustor vibration consistency evaluation system based on multi-modal fusion, which is used to implement the above-mentioned gas turbine combustor vibration consistency evaluation method based on multi-modal fusion. The system includes: Feature fusion module: used to obtain the multi-dimensional historical combustion vibration data set of the gas turbine combustor; the historical combustion vibration data set includes the vibration signals and temperature signals of M combustors; perform segmented processing on the historical combustion vibration data set to form n1 sub-datasets; for each sub-dataset, construct the preliminary feature vector based on the Mahalanobis distance, and construct the multi-modal feature vector based on mutual information weighted fusion according to the preliminary feature vector; Criteria generation module: used to build a deep feature extraction model to extract deep features of multimodal feature vectors; perform cluster analysis on deep features to classify the health status level of combustion chamber vibration; calculate the cluster center of each health status level to form a multi-level health status level criterion;

[0018] Health level classification module: used to collect the real-time status signal of each combustion chamber under the current operating state, and determine the health status level of the current combustion chamber vibration based on the real-time status signal and multi-level health status level criteria.

[0019] An electronic device comprises a memory, a central processing unit and a computer program stored in the memory and executable on the central processing unit, wherein the central processing unit implements the above-mentioned gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion when executing the computer program.

[0020] A computer-readable storage medium stores a computer program, which, when executed, implements the above-mentioned gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] By fusing the multimodal data of vibration signals and temperature signals and combining deep learning to extract deep features, the vibration state of the combustion chamber can be characterized more comprehensively and accurately, and the accuracy and reliability of the evaluation can be improved. By using adaptive operating condition segmentation and clustering methods, different operating conditions can be automatically divided according to actual operating data, so that the evaluation method can adapt to the changes in vibration characteristics under different operating conditions and has strong adaptability. By constructing a multi-level health status grade criterion, real-time dynamic evaluation of the vibration state of the combustion chamber can be achieved, and potential problems can be discovered in time. By accurately evaluating the vibration consistency of the combustion chamber, abnormal conditions can be discovered early, effectively reducing the operating risks of the gas turbine and improving equipment safety. Accurate evaluation of the vibration state helps to optimize maintenance strategies, avoid unnecessary shutdowns and overhauls, and extend the service life of the gas turbine. By accurately judging the vibration state, predictive maintenance can be achieved, unplanned downtime can be reduced, and maintenance costs can be reduced. Accurate evaluation of vibration consistency helps to optimize the working state of the combustion chamber, improve combustion efficiency, and thus improve the overall efficiency of the gas turbine. Through automated data processing and evaluation processes, the subjectivity of human judgment is reduced, and the objectivity and consistency of the evaluation are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 It is a principle flow chart of the gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion in the present invention; Figure 2 A flow chart of a method for segmenting a historical combustion vibration data set in a gas turbine combustor vibration consistency assessment method based on multi-modal fusion of the present invention; Figure 3 A flow chart of a method for constructing a preliminary feature vector based on Mahalanobis distance in a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion of the present invention; Figure 4 A flow chart of a method for constructing a multi-modal feature vector based on mutual information weighted fusion in a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion of the present invention; Figure 5 A flow chart of a method for extracting deep features of multi-modal feature vectors in a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion according to the present invention; Figure 6 A flow chart of a method for forming a multi-level health status grade criterion in a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion according to the present invention; Figure 7 A flow chart of a method for determining the operating section to which the current operating state belongs in the gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion of the present invention; Figure 8 It is a functional module diagram of the gas turbine combustion chamber vibration consistency assessment system based on multi-modal fusion in the present invention. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] Example 1

[0027] See also Figure 1As shown, this embodiment provides a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion, including: Step S1000, obtaining a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; the historical combustion vibration data set includes a vibration signal and a temperature signal; segmenting the historical combustion vibration data set to form n1 sub-data sets; for each sub-data set, constructing a preliminary feature vector based on Mahalanobis distance, and constructing a multi-modal feature vector based on mutual information weighted fusion according to the preliminary feature vector; Furthermore, step S1000 includes: Step S1100, obtaining a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; M is a positive integer; Specifically, in order to obtain data that fully reflects the state of the combustion chamber, it is necessary to collect vibration and temperature signals in detail. Select a piezoelectric acceleration sensor suitable for high-temperature environments, and arrange multiple sensors on the outer surface of the flame tube end cover of each combustion chamber, evenly distributed along the circumference and perpendicular to the axis of the flame tube; ensure that each combustion chamber can capture vibration signals. The sensor collects vibration acceleration signals from the combustion chamber structural parts to reflect the vibration state during the combustion process. Collect vibration signals from each combustion chamber, including acceleration, frequency and other information. The high temperature and high pressure environment of the combustion process places strict requirements on the sensor. The selected piezoelectric acceleration sensor has good high temperature resistance and frequency response characteristics, and is highly reliable. Arrange the acceleration sensor at the key part of the flame tube-the end cover, which bears the cantilever load of the hot end components and has obvious vibration response. By adopting a multi-point arrangement, the circumferential and axial vibrations of the combustion chamber can be collected to obtain spatial distribution characteristics.

[0028] Select a platinum resistance temperature sensor, and extend the temperature sensor into the combustion chamber through a special installation sleeve. The temperature sensor uses a platinum resistor with high precision and good stability, a special high-temperature sleeve and a reasonable burial depth to ensure the reliable measurement of the temperature probe in the high-temperature airflow. Arrange multiple temperature sensors at the head and tail respectively to obtain the temperature distribution along the process, providing a basis for the subsequent analysis of the non-uniformity of the hot end temperature field. The high-frequency sampling and continuous acquisition of the data acquisition instrument can obtain the dynamic characteristic data of the combustion chamber at different times, providing data guarantee for mining deep-level time-frequency domain characteristics. The reasonable arrangement of sensors and matching of high-quality data acquisition equipment ultimately form a comprehensive and detailed multi-dimensional time series data set, which is an important basis for vibration consistency evaluation.

[0029] Step S1200, segmenting the historical combustion vibration data set to form n1 sub-data sets; Furthermore, if Figure 2 As shown, step S1200 includes: Step S1210, extracting gas turbine operating parameters corresponding to the historical combustion vibration data set, wherein the operating parameters include load and speed; Step S1220, set the load change threshold δ, perform a traversal analysis on the historical combustion vibration data set in time series, take time t0 as the starting point, if the load change at time t1 relative to the starting time t0 exceeds the load change threshold δ, it is considered that the operating condition has switched, and mark time t1 as the operating condition switching point; then take time t1 as the new starting point, repeat the above process until the end of the historical combustion vibration data set is traversed; Step S1230, dividing the historical combustion vibration data set into n2 initial operating condition segments according to the operating condition switching point; Step S1240, clustering the operating condition parameters in the initial operating condition segment, and dividing the historical combustion vibration data set into n1 subdivided operating condition segments; Further, step S1240 includes: Step S1241, using the operating parameters in the initial operating condition section as sample features to form a feature matrix X, the dimension of the feature matrix X is m×n, m is the number of samples, and n is the number of features; Step S1242, set the value range of the cluster number K to ,in Indicates rounding down; traverse the cluster number K value and calculate the silhouette coefficient sequence and DBI index sequence; Step S1243, calculating a comprehensive metric index sequence according to the silhouette coefficient sequence and the DBI index sequence, and taking the number of clusters corresponding to the maximum value in the comprehensive metric index sequence as the optimal number of clusters N1; Step S1244, fix the optimal cluster number N1, use the K-Means algorithm to cluster the feature matrix X, obtain n1 subdivided operating condition segments and the cluster center of each subdivided operating condition segment, and form the cluster center vector {C1, C2, ..., C n1}; where N1=n1, C1 is the cluster center of the first subdivided operating condition segment, C2 is the cluster center of the second subdivided operating condition segment, and C n1 It is the cluster center of the n1th subdivided operating condition segment.

[0030] Step S1250: Generate a corresponding sub-data set for each subdivided operating condition segment to obtain n1 sub-data sets.

[0031] Specifically, in actual operation, a gas turbine undergoes processes such as startup, shutdown, and variable operating conditions. The vibration characteristics of the combustion chamber under different operating conditions are significantly different. To improve the pertinence of subsequent evaluations, it is necessary to separate the data under different operating conditions. First, key parameters reflecting the state of the gas turbine, such as load and rotational speed, are extracted based on the operation records. The load refers to the active power of the generator set driven by the gas turbine and is a key parameter reflecting the output level of the gas turbine. The rotational speed refers to the rotational speed of the gas turbine rotor, usually expressed in revolutions per minute (RPM), and is another important parameter reflecting the operating state of the gas turbine. By analyzing the operation records, a series of load and rotational speed data corresponding to the historical vibration data can be obtained, providing a basis for the division of operating conditions.

[0032] Considering the fluctuations in load and measurement errors in actual operation, it is not that a minor change in load is considered a change in operating conditions. Instead, a threshold value δ for load change is set. Only when the load change amount exceeds this threshold value is it considered that a significant change in operating conditions has occurred, and it is marked as an operating condition switching point. The threshold value δ can be set according to experience, such as 5% of the rated load. Taking load as an example, when the load change exceeds the preset threshold value, such as 5%, it can be determined as a significant change in operating conditions. Based on this until the end of the data, several switching moments can be obtained. Segmenting the data set with the switching moments as the boundary can initially achieve the separation of data under different operating conditions.

[0033] Segmenting the data according to the load change threshold δ results in n2 initial operating condition segments. This division method mainly considers the impact of sudden changes in load and cuts the data with the load switching point as the boundary. However, in actual operation, even if the load change is small, changes in other operating condition parameters, such as rotational speed, may also lead to differences in the combustion state. To further refine the division of operating conditions, it is necessary to comprehensively consider the similarity of various operating condition parameters, which requires performing clustering analysis on the basis of the initial segmentation. Clustering analysis is an unsupervised learning method that can automatically divide data into different groups according to the similarity of sample features, making the similarity within the group high and the difference between groups large. Using the operating condition parameters such as load and rotational speed within each initial operating condition segment as sample features, through clustering, samples with closer operating conditions can be automatically grouped into one category, obtaining a more detailed division of operating conditions. The number of refined operating conditions n1 formed by clustering is usually determined by the distribution characteristics of the data itself and may be greater than or less than the initial segmentation number n2. When the sample features of different initial operating condition segments are very close, they are likely to be grouped into the same category, resulting in n1 < n2; on the contrary, when there are obvious differences within the same initial operating condition segment, it may be subdivided into multiple operating conditions, resulting in n1 > n2.

[0034] The load, speed and other operating parameters in the initial operating section are used as sample features to form a feature matrix X, and the feature matrix X is subjected to necessary dimensionless processing. The multivariate features of each sample are organized into a matrix form to facilitate unified algebraic operations and transformations. Different features often have different dimensions and numerical ranges. In order to eliminate the impact of such differences, the features need to be dimensionless, such as maximum and minimum value normalization, Z-score standardization, etc. The silhouette coefficient measures the similarity between the sample and its class and the difference between the sample and the nearest class. The value range is [-1,1], and the larger the better. The DBI value range is [0,+∞), which comprehensively considers the compactness within the cluster and the separation between classes. The smaller the DBI value, the better the clustering effect. Set the K value range to [2, ], traverse each K value, calculate the corresponding silhouette coefficient and DBI index, and select the optimal number of clusters N1. In order to comprehensively consider the silhouette coefficient and DBI index, a combined metric function can be constructed to perform a weighted average of the two indicators. The weight can be adjusted according to actual needs to calculate the comprehensive metric index; the sum of the weights of the two indicators is equal to 1. The larger the weight of the silhouette coefficient, the more attention is paid to the compactness of the clusters, and the larger the weight of the DBI index, the more attention is paid to the separation of the clusters. The number of clusters when the comprehensive metric index is the largest is selected as the optimal number of clusters.

[0035] The optimal number of clusters N1 is fixed, and the K-Means algorithm is used to perform clustering training on the feature matrix X. Initialize N1 cluster centers, minimize the cost function through iterative optimization, and continuously update the cluster centers and sample category labels until convergence. Each category in the clustering results is treated as an independent working condition, forming a total of n1 subdivided working conditions, n1=N1. Extract the time period contained in each subdivided working condition, summarize the vibration and temperature data in these time periods, and generate the sub-dataset corresponding to the working condition.

[0036] Through the above clustering steps, the initial n2 rough working condition segments are finally refined into n1 more detailed working condition categories, and each category forms a sub-dataset. This clustering subdivision can characterize the similarity of working conditions from the high-dimensional feature space as a whole, making up for the shortcomings of single feature division, making the final working condition subset more accurate and reliable. Through the silhouette coefficient and DBI index, the clustering quality under different clustering numbers can be quantitatively evaluated, providing an objective basis for working condition clustering. Compared with manually setting the number of clusters, this data-driven optimization method can adaptively discover the intrinsic cluster structure of the data, avoiding subjectivity and blindness.

[0037] Clustering and subdividing working conditions improves the pertinence of subsequent modeling. Different working conditions have different data characteristics, and the factors that affect the combustion state are different. After subdividing the working conditions, the unique influencing mechanism can be explored for each working condition, which helps to improve the accuracy of the evaluation model. Reduce the complexity of modeling. If all working condition data are modeled together, the model needs to have strong expression ability to characterize complex data distribution, which often comes at the cost of model complexity. After subdividing the working conditions, the data distribution in each sub-condition is relatively simple, the requirements for model capabilities are reduced, and it is easier to obtain an ideal fitting effect. It is conducive to knowledge accumulation. The natural division of working conditions is obtained through clustering, which reflects the inherent laws of the actual operation of the gas turbine. By analyzing the characteristics of different working conditions, we can summarize their commonalities and individualities, formulate operation and maintenance strategies in a targeted manner, and accumulate valuable engineering experience.

[0038] For example, the historical data of a gas turbine is initially divided into 5 operating conditions, and then the load, speed and other data in each operating condition are extracted and normalized to form a feature matrix with 1000 samples and 4 features. Then the K-Means algorithm is used to cluster the matrix, with K set to 2~10, and the contour coefficient and DBI index are calculated. The results show that when K=7, the contour coefficient reaches a maximum value of 0.68 and the DBI index reaches a minimum value of 1.21, so the optimal number of clusters is determined to be 7. All samples are divided into 7 categories, each corresponding to a subdivided operating condition, and all vibration and temperature data in these 7 operating conditions are extracted to form 7 sub-data sets.

[0039] Step S1300, constructing a preliminary feature vector based on Mahalanobis distance for each sub-data set; Furthermore, if Figure 3 As shown, step S1300 includes: Step S1310: for each sub-data set, calculate the Mahalanobis distance of each combustion chamber vibration signal relative to the mean of the sub-data set to form a vibration Mahalanobis distance vector ,in, represents the vibration Mahalanobis distance vector of combustion chamber j in the i-th operating condition, i represents the operating condition number, and j represents the combustion chamber number; Step S1320: for each sub-data set, calculate the Mahalanobis distance of each combustion chamber temperature signal relative to the mean of the sub-data set to form a temperature Mahalanobis distance vector ;in, represents the temperature Mahalanobis distance vector of combustion chamber j in the i-th operating condition; Step S1330: The vibration Mahalanobis distance vector and the temperature Mahalanobis distance vector are combined into a preliminary feature vector according to the combustion chamber sequence number. , , Represents the preliminary characteristic vector of combustion chamber j in the i-th operating condition.

[0040] Specifically, in step S1300, for each sub-data set, a reasonable feature needs to be designed to characterize the degree to which the state of each combustion chamber deviates from the normal value. Mahalanobis distance is an effective anomaly detection method. By calculating the weighted distance from the sample point to the feature mean, the degree to which the point deviates from the distribution center can be measured. First, for the vibration data of each combustion chamber, the Mahalanobis distance between it and the mean of all samples in the sub-data set is calculated to form a one-dimensional feature vector reflecting the degree of vibration deviation. Similarly, for the temperature data of each combustion chamber, its Mahalanobis distance feature vector is also calculated. The vibration Mahalanobis distance and temperature Mahalanobis distance corresponding to the same combustion chamber are combined to finally form a two-dimensional feature vector. This feature vector characterizes the degree of abnormality of different combustion chambers under various operating conditions from the two aspects of vibration and temperature, providing a judgment basis for subsequent vibration consistency evaluation. Compared with traditional distance metrics such as Euclidean distance, Mahalanobis distance takes into account the correlation between different state parameters. The larger the distance value, the higher the possibility that the sample point deviates from the normal operating condition. Combining the Mahalanobis distances of vibration and temperature into a unified feature vector can more comprehensively reflect the health status of the combustion chamber and avoid misjudgment caused by a single physical quantity. It can be seen that the construction of Mahalanobis distance features can effectively screen out abnormal data that deviates from normal operating conditions, and is an important means to achieve preliminary fault detection.

[0041] Step S1400 , for each sub-data set, construct a multimodal feature vector based on mutual information weighted fusion according to the preliminary feature vector.

[0042] Furthermore, if Figure 4 As shown, step S1400 includes: Step S1410: For each combustion chamber, calculate the mutual information matrix between the vibration signal and the temperature signal ; represents the mutual information matrix between the vibration signal and the temperature signal in the i-th operating condition; Step S1420: Mutual information matrix Perform normalization to obtain the fusion weight matrix ; represents the fusion weight matrix of the i-th operating condition; Step S1430: According to the fusion weight matrix Weighted fusion of the i-th operating condition segment The vibration Mahalanobis distance vector of the combustion chamber and temperature Mahalanobis distance vector , construct multimodal fusion features ; Indicates the i-th operating condition segment Multi-modal fusion characteristics of a combustion chamber; Step S1440: Fusion of multi-modal features of each combustion chamber Composition Multi-mode feature vector of each operating condition , , M is the number of combustion chambers, Indicates The multi-modal characteristic vector of the first combustion chamber in each operating section, Indicates The multi-modal characteristic vector of the second combustion chamber in the operating section, Indicates The multi-modal characteristic vector of the Mth combustion chamber in the operating section.

[0043] Specifically, in step S1400, in order to fully explore the intrinsic relationship between the vibration signal and the temperature signal, a multimodal fusion feature based on mutual information weighted fusion is further constructed. Mutual information can measure the correlation between two random variables. The larger the mutual information value, the stronger the correlation between the two variables. First, the mutual information between the vibration signal and the temperature signal in each operating section is calculated to form a mutual information matrix. The elements in the matrix Describes the The accelerometer and To facilitate subsequent fusion, the mutual information matrix is ​​normalized so that the sum of its elements is 1, and the fusion weight matrix is ​​obtained. The normalized weights reflects the acceleration-temperature sensor pair ( , ) on this basis, the vibration Mahalanobis distance vector of each combustion chamber is and temperature Mahalanobis distance vector Perform weighted fusion. Each component and Each component of is combined in pairs and then multiplied by the corresponding fusion weight , and finally superimpose and sum to obtain the scalarized multimodal fusion features The fusion features of all combustion chambers are summarized to finally form the multi-modal feature vector of this operating section. . The fusion feature integrates the complementary information of vibration and temperature, overcomes the limitations of a single mode, and can more comprehensively reflect the state of the combustion chamber. The introduction of mutual information-driven fusion weights allows sensor pairs with strong correlation to obtain higher weights in the fusion, highlighting the role of fault-related modes, which has physical significance. The adaptive weight calculation method can adjust the weights in real time according to the dynamic changes of the data, thereby improving the robustness of the algorithm. Multimodal fusion reduces the redundancy and interference of information, obtains a more refined and high-level state representation, and is the key to achieving high-precision vibration consistency assessment.

[0044] Step S2000, constructing a deep feature extraction model to extract deep features of the multimodal feature vector; performing cluster analysis on the deep features to classify the health status level of the combustion chamber vibration; calculating the cluster center of each health status level to form a multi-level health status level criterion; the health status level of the combustion chamber vibration includes normal, warning and abnormal; Furthermore, step S2000 includes: Step S2100, constructing a deep feature extraction model to extract deep features of the multimodal feature vector; Furthermore, if Figure 5 As shown, step S2100 includes: Step S2110, the deep feature extraction model includes a convolutional neural network and an autoencoder, the multimodal feature vector is input into the convolutional neural network, and a high-dimensional feature map is extracted; Step S2120, inputting the high-dimensional feature map output by the convolutional neural network into the autoencoder to learn global compression low-dimensional features; Step S2130, using the globally compressed low-dimensional features output by the autoencoder as deep features of the multimodal feature vector.

[0045] Specifically, in step S2100, a combined structure of a convolutional neural network and an autoencoder is used to construct an end-to-end deep feature extraction model. First, the multimodal feature vector obtained in step S1400 is input into the convolutional neural network. Through local connections and weight sharing, the convolutional neural network can efficiently extract local features of the data. Through operations such as convolution and pooling, the convolutional neural network converts the initial multimodal fusion features into increasingly abstract feature maps layer by layer. The shallower feature maps extract low-order statistics of vibration and temperature signals, while the deeper feature maps contain more complex and global features. The advantage of the convolutional neural network is that the data is gradually abstracted through a hierarchical structure, and the most valuable key features for state discrimination are automatically learned. These local features comprehensively reflect the detailed information of the combustion state.

[0046] Then, the high-dimensional feature map output by the convolutional neural network is input into the autoencoder network. The autoencoder is an unsupervised feature learning method consisting of two parts: an encoder and a decoder. The encoder maps the input data to a low-dimensional feature space, and the decoder attempts to reconstruct the original data from the compressed features. By minimizing the reconstruction error, the autoencoder can learn an efficient compressed representation of the input data, eliminating redundancy and noise in the data. In this step, the encoder compresses the convolutional feature map to a lower dimension and extracts the global features of the data. These global features summarize the main patterns of the combustion state in the most concise way. The decoder is used to constrain the encoder to learn meaningful features, making the compressed features well interpretable.

[0047] Finally, the low-dimensional features output by the autoencoder are used as the deep feature representation of the multimodal feature vector. These deep features inherit the local feature extraction capability of the convolutional neural network and the global compression capability of the autoencoder, and can fully characterize the inherent laws of the combustion chamber vibration. Let M represent the number of combustion chambers and S represent the dimension of the deep feature. Then the dimension of the deep feature F output by step S2100 is M×S.

[0048] The construction of the deep feature extraction model is mainly based on data-driven end-to-end learning, and the model parameters are iteratively optimized through optimization algorithms such as stochastic gradient descent. Given a large number of multimodal fusion feature samples, the model can adaptively learn the optimal feature transformation from the data by minimizing loss functions such as reconstruction error. This data-driven learning paradigm overcomes the limitations of manually designed features, does not rely on expert experience and prior knowledge, and can discover the key patterns inherent in the data. By extracting local features through convolutional neural networks and learning global compression features with autoencoders, the multi-level features of the combustion state can be characterized at different scales, which not only retains detailed information but also reveals global laws. Compared with traditional artificial features, the features extracted by deep learning are often more discriminative, robust, and adaptable, laying a solid foundation for subsequent state clustering.

[0049] In summary, step S2100 uses a deep learning model to extract highly refined deep features from multimodal fusion features. These deep features summarize the intrinsic patterns of combustion chamber vibration in the most concise and discriminative way, reflecting the advantages of learning from multi-source data. Deep feature extraction is the key to knowledge discovery and intelligent decision-making, and directly determines the accuracy of subsequent state clustering. The state criteria constructed based on deep features can quickly and accurately assess the health level of the combustion chamber and guide operation and maintenance personnel to take targeted maintenance measures. It can be seen that deep feature learning is of great significance in combustion chamber vibration monitoring and represents a new direction for intelligent fault diagnosis and health management.

[0050] Step S2200, cluster analysis is performed on the deep features of the multimodal feature vector to classify the health status levels of the combustion chamber vibration; the health status levels of the combustion chamber vibration include normal, warning and abnormal.

[0051] Specifically, in step S2200, the deep feature vector is divided into several state clusters using the cluster analysis method, and these clusters are further classified into three health status levels. Cluster analysis is a typical unsupervised learning method that does not require prior labeling of data, but instead achieves automatic data division by maximizing intra-cluster similarity and inter-cluster differences. In this step, K-means, as the most classic clustering algorithm, is used to characterize the distribution pattern of deep features. K-means can discover natural cluster structures in the feature space by optimizing the sum of squared distances between samples and cluster centers. By iteratively optimizing cluster centers and cluster properties of samples, K-means can discover meaningful state patterns from deep features, and each cluster corresponds to a typical combustion chamber working state.

[0052] A key issue of the K-means algorithm is the selection of the number of clusters. Different numbers of clusters will lead to different clustering results. If the number of clusters is too small, some important structures of the data may be ignored. If the number of clusters is too large, samples with similar nature may be divided into different clusters, reducing the generalization of clustering. In practical applications, it is often necessary to optimize the number of clusters. In step S2220, the silhouette coefficient and Davies-Bouldin Index (DBI) index are used as evaluation indicators of clustering performance. The silhouette coefficient measures the similarity of the sample to the cluster to which it belongs and the difference from other clusters. The value is between -1 and 1. The larger the value, the better the clustering effect. The DBI index comprehensively considers the compactness within the cluster and the degree of separation between clusters. The smaller the value, the higher the clustering quality. By optimizing the silhouette coefficient and DBI index by cross-validation, the optimal number of clusters can be selected from the data, which will neither overfit local features nor underfit global structures.

[0053] After obtaining the optimal clustering results, these clusters need to be further interpreted as different health status levels. Based on expert knowledge and historical experience, clusters can be divided into three types: normal, warning, and abnormal. The combustion chamber near the center of the normal cluster has the most stable working state, and its vibration characteristics are close to most samples, representing a healthy baseline state. Clusters that are far away from the center of the normal cluster but have not yet reached the level of failure have vibration levels that are higher than normal, but are still within the controllable range and are marked as warning states. For samples that are far away from the normal cluster and close to the center of the abnormal cluster, their vibration patterns are obviously abnormal and can be judged as a fault state, requiring timely maintenance. By reasonably dividing the status levels, a refined assessment of the health level of the combustion chamber can be achieved, providing a quantifiable basis for status monitoring, fault warning, and maintenance decisions.

[0054] In summary, step S2200 constructs state clusters that reflect the typical working conditions of the combustion chamber in the deep feature space through cluster analysis and state division. These clusters summarize the key patterns in massive historical data in a data-driven manner, overcoming the subjectivity of manually defining state boundaries. Through the interpretation and association of state clusters, complex sensor signals can be intuitively mapped to health levels that are easy to make decisions, making state monitoring more intelligent. It is worth mentioning that the cluster analysis in this step is performed on the basis of deep features. The combination of deep feature learning and state clustering can maximize the intrinsic value of the data and form an interpretable and operational combustion chamber health assessment system. The multi-level state division based on clustering provides important support for subsequent hierarchical diagnosis and predictive maintenance, and represents a new way of equipment health management.

[0055] Step S2300, calculating the cluster center of each health status level to form a multi-level health status level criterion; Furthermore, if Figure 6 As shown, step S2300 includes: Step S2310, respectively calculating the coordinates Cn, Cw, Ca of the cluster centers of the three health status levels of normal, warning, and abnormal; Step S2320, calculating the distance between each sample in the normal state cluster and the normal state cluster center coordinate Cn, and recording the maximum distance as d1; Step S2330, calculating the distance between each sample in the warning state cluster and the center coordinate Cn of the normal state cluster, and recording the minimum distance as d2; Step S2340, set δ1=1 / 2(d1+d2) as the boundary threshold of the normal state; Step S2350, calculating the distance between each sample in the warning state cluster and the coordinate Cw of the center of the warning state cluster, and recording the maximum distance as d3; Step S2360, calculating the distance between each sample in the abnormal state cluster and the center coordinate Cw of the warning state cluster, and recording the minimum distance as d4; Step S2370, set δ2=1 / 2(d3+d4) as the boundary threshold of the warning state; Step S2380, calculate the distance from each sample in the abnormal state cluster to the abnormal state cluster center coordinate Ca, record the maximum distance as d5, and set δ3=2×d5 as the boundary threshold of the abnormal state.

[0056] Specifically, the core of step S2300 is to use the clustering method of unsupervised learning to automatically discover the intrinsic cluster structure of samples in the deep feature space, and use the cluster center as a typical representative of the health state, and the cluster boundary as the criterion threshold for state division. K-means clustering is a typical distance-based partitioning clustering algorithm, which achieves clustering by minimizing the sum of squares of the distances from the samples to the cluster centers. When clustering, the Euclidean distance between each sample and the cluster center is calculated, and the samples are divided into the clusters with the closest distance. When the center is updated, the geometric center coordinates of the samples in each cluster are recalculated as the new cluster center. The above process is repeated continuously, so that the cluster center gradually stabilizes and can better represent the characteristic distribution of each type of sample.

[0057] This step clusters the historical samples in the deep feature space into three categories, corresponding to the three health states of the combustion chamber: normal, warning, and abnormal. In the normal state, the parameters of the combustion chamber are stable and the operating conditions are close to most samples; in the warning state, the combustion chamber begins to fluctuate slightly, but has not yet reached the abnormal level; in the abnormal state, the combustion chamber deviates significantly from the normal operating conditions and requires attention. These three typical states are automatically identified through clustering, and the coordinates of their cluster centers are used as representative features of the state. When a newly collected sample is close to a cluster center, it can be judged that the sample is likely to be in the healthy state represented by the center.

[0058] In order to quantitatively divide the boundaries of different states, it is necessary to estimate appropriate thresholds based on clustering. The boundary threshold δ1 of the normal state takes the average of the maximum distance d1 within the normal cluster and the minimum distance d2 from the warning cluster to the normal center, that is, samples falling within the radius of δ1 tend to be judged as normal. The boundary threshold δ2 of the warning state takes the average of the maximum distance d3 within the warning cluster and the minimum distance d4 from the abnormal cluster to the warning center. The boundary threshold δ3 of the abnormal state takes twice the maximum distance d5 within the abnormal cluster. Samples with a distance from the abnormal center exceeding δ3 often mean serious faults. Based on the clustering results, the above threshold selection method fully considers the compactness within the cluster and the separation between classes, and can reasonably delineate the effective boundaries of different states.

[0059] This judgment method does not require manual setting of state thresholds, but automatically learns the distribution characteristics of typical states based on the clustering structure of historical data, which is adaptive; secondly, using the cluster center as a state model can reduce the impact of individual extreme samples and improve the robustness of the judgment; thirdly, the fusion takes into account the compactness within the cluster and the degree of separation between clusters, which can better balance the classification accuracy and generalization performance; finally, by comparing the distances to the three cluster centers, the fuzzy interval caused by a single threshold division can be avoided, making the state judgment more refined.

[0060] Step S3000, collecting the real-time status signal of each combustion chamber under the current operating state, and determining the health status level of the current combustion chamber vibration according to a multi-level health status level criterion.

[0061] Furthermore, step S3000 includes: Step S3100, collecting real-time status signals of each combustion chamber under the current operating state, wherein the real-time status signals include real-time vibration signals and real-time temperature signals; determining the operating section to which the current operating state belongs, and marking it as the current operating section; and calculating the real-time multi-modal feature vector of the real-time status signal; Furthermore, if Figure 7 As shown, step S3100 includes: Step S3110, extracting real-time operating parameters under the current operating state; Step S3120: perform dimensionless processing on the real-time operating condition parameters and convert the real-time operating condition parameters into a real-time operating condition vector x (c) ; Step S3130, calculate the cluster center vector {C1, C2, ..., C n1}, and obtain the distance vector ; represents the Euclidean distance between the real-time operating condition vector and the cluster center of the first subdivided operating condition segment, represents the Euclidean distance between the real-time operating condition vector and the cluster center of the second subdivided operating condition segment, Represents the Euclidean distance between the real-time operating condition vector and the cluster center of the n1th subdivided operating condition segment; Step S3140: The real-time operating condition vector x (c) Divide into the subdivided operating condition segment corresponding to the nearest cluster center as the operating condition segment of the current operating state; Step S3150, calculating the Mahalanobis distance of the real-time vibration signal relative to the mean of the sub-data set of the corresponding operating section to form a real-time preliminary feature vector; Step S3160, according to the fusion weight matrix and real-time preliminary feature vectors, and construct real-time multimodal feature vectors based on mutual information weighted fusion.

[0062] Specifically, step S3100 is the online monitoring stage, which senses the changes in the vibration state of the combustion chamber in real time. To this end, it is necessary to arrange piezoelectric acceleration sensors and platinum resistance temperature sensors with the same configuration as in the offline training stage on the unit. The installation location, quantity, model specifications, etc. of the sensors must be consistent with those in the offline stage to ensure the comparability of the collected data.

[0063] In addition, it is also necessary to automatically determine the type of operating condition of the current signal based on the real-time operating parameters of the gas turbine, such as load and speed, using the operating condition classification rules summarized in the offline stage, so as to match the corresponding fusion weights and evaluation thresholds for subsequent feature calculations. First, extract the key parameters that reflect the current state of the gas turbine, such as load and speed, to form real-time operating parameters. Since the dimensions and numerical ranges of these parameters may differ from the training data in the offline stage, they need to be dimensionlessly processed to uniformly map each component to the interval [0,1] to obtain the real-time operating vector x (c) The specific method of dimensionless processing needs to be consistent with the offline stage to ensure that the vector x (c) The cluster center vectors of each subdivided working condition are compared in the same scale space. In the offline stage, the K-Means algorithm is used to obtain n1 cluster center vectors (see step S1244), each of which represents a typical position of a subdivided working condition in the high-dimensional feature space. Therefore, the vector x can be calculated. (c) The Euclidean distance from each cluster center is used to measure the closeness between the current working condition and each subdivided working condition. The smaller the distance, the more similar the characteristic distribution of the current working condition parameters is to the subdivided working condition. Traverse all n1 subdivided working conditions and calculate their distance from x (c) The Euclidean distance of . Finally, the subdivided working condition corresponding to the nearest cluster center is selected as the working condition segment to which the current operating state belongs. Once the working condition segment to which the state belongs is determined, the evaluation model for the working condition established in the offline stage can be called to perform an online evaluation of the current vibration consistency level. It can be seen that the above working condition matching process avoids the secondary clustering of real-time data, makes full use of the working condition division knowledge summarized in the offline stage, and realizes rapid working condition identification by comparing the distance between the normalized vector and the cluster center, which greatly reduces the calculation complexity of the online stage. At the same time, the use of Euclidean distance, a simple and intuitive similarity measure, is easy to implement in engineering, and corresponds to the clustering method in the offline stage to maintain consistency. In practical applications, the Euclidean distance can also be weighted and corrected to highlight the influence of certain important working condition parameters and further improve the accuracy of working condition division. In general, step S3200 provides an effective way to achieve real-time working condition division based on clustering results, laying the foundation for subsequent feature extraction, consistency evaluation and other links.

[0064] Then, using the feature fusion method given in steps S1300 and S1400, calculate the real-time multi-modal feature vector of the currently collected real-time status signal. First, calculate the Mahalanobis distances of the real-time vibration signals and real-time temperature signals of each combustion chamber with respect to the mean of the sub-dataset of the corresponding working condition section respectively to form a preliminary feature vector. Then, use the mutual information fusion weight matrix calculated in the offline stage to perform weighted fusion on the vibration features and temperature features in the real-time preliminary feature vector to obtain the real-time multi-modal fusion features of each combustion chamber under the current operating state. Finally, combine the multi-modal fusion features of each combustion chamber in sequence to form a multi-modal feature vector representing the current operating state of the entire combustion system. This series of calculations can be automatically completed with the help of a trained deep learning model to achieve rapid characterization and evaluation of the combustion state.

[0065] Through the preprocessing and feature fusion of real-time signals, the obtained multi-modal feature vector can comprehensively reflect the dynamic changes of the combustion chamber vibration. On the one hand, the sensors are reasonably arranged and the data collection is strictly implemented to ensure the consistency and comparability of the monitoring data and the training data; on the other hand, the working condition is automatically judged and the fusion weights are dynamically loaded, endowing the online evaluation with the ability of working condition adaptability, ensuring the reliability and continuity of the evaluation process. Input the monitoring data into the trained deep learning model for inference, and it can quickly calculate the degree to which the current operating state deviates from the normal mode, timely detect the abnormal signs of the combustion system, and provide a basis for fault warning and operation and maintenance decision-making.

[0066] Step S3200: Input the real-time multi-modal feature vector into the deep feature extraction model to obtain the real-time deep features; calculate the Euclidean distances Dn, Dw, Da between the real-time deep features and the coordinates Cn, Cw, Ca of the clustering centers of the three health state levels of normal, warning, and abnormal, and determine the health state level of the current combustion chamber vibration; If Dn ≤ δ1 and Dw ≤ Da, it is determined to be in the normal state; If δ1 < Dn ≤ δ2 and Dw ≤ Da, it is determined to be in the warning state; If δ2 < Dn ≤ δ3, it is determined to be in the abnormal state; If Dn > δ3, it is determined to be in the severe fault state.

[0067] Specifically, in step S3200, the real-time collected combustion chamber vibration and temperature signals are subjected to the same feature construction steps as the offline stage to form a real-time multimodal feature vector. Then the real-time multimodal feature vector is input into the trained deep feature extraction model to obtain the real-time deep feature after dimensionality reduction. The real-time deep feature is a high-level abstraction of the original high-dimensional feature, which can better reflect the essential characteristics of the data. In order to judge the health status level of the current combustion chamber, it is necessary to calculate the Euclidean distance between the real-time deep feature and the cluster center of the three health levels. The Euclidean distance is a commonly used similarity metric. The smaller the distance, the closer the feature is to the center of the class. Cn, Cw, and Ca represent the coordinates of the cluster centers of the three levels of normal, warning, and abnormal, respectively, and Dn, Dw, and Da represent the Euclidean distance from the real-time deep feature to the three centers. δ1, δ2, and δ3 are the boundary thresholds of the three levels.

[0068] If the distance Dn from the real-time depth feature to the normal center does not exceed δ1, and the distance Dw to the warning center is less than the distance Da to the abnormal center, it can be determined that the current combustion chamber is in a normal state. This means that the current feature is highly consistent with the normal mode, the combustion chamber is running smoothly, and all indicators are within a reasonable range.

[0069] If the real-time depth feature deviates from the normal center by more than δ1 but less than δ2, and the distance to the warning center Dw is less than the distance to the abnormal center Da, it can be judged as a warning state. In this case, although the combustion chamber has a certain degree of abnormality, it has not yet reached the point where it seriously affects safety. It is worth paying attention to and strengthening monitoring, but generally no emergency shutdown is required.

[0070] If the real-time depth feature deviates from the normal center by more than δ2 but less than δ3, it can be judged as an abnormal state. This means that the combustion chamber has obvious abnormalities. Although it has not caused permanent damage, it may deteriorate into a fault if not handled in time. Close attention should be paid, and combined with other information for comprehensive analysis, necessary measures should be taken in a timely manner to prevent the situation from escalating.

[0071] If the distance Dn of the real-time depth feature from the normal center exceeds the maximum threshold δ3, it is highly suspected that a serious fault has occurred. At this time, the state of the combustion chamber is completely beyond the normal range. If emergency measures are not taken immediately, it is very likely to cause an accident, resulting in serious consequences such as equipment damage and safety accidents.

[0072] The above discrimination method fully draws on the idea of ​​pattern recognition, matches the real-time features with the standard pattern through distance measurement, and automatically identifies the closest health level. The purpose of setting the distance threshold is to divide the severity of different levels. δ1, δ2, and δ3 can be regarded as the boundaries of the three levels of normal, warning, and abnormal. The setting of multi-level thresholds can more carefully depict the evolution of the health status, detect abnormal signs early, and provide a basis for subsequent maintenance decisions.

[0073] Compared with the traditional single threshold discrimination, this health assessment method can more comprehensively and meticulously characterize the state of the combustion chamber, and achieve "graded warning and early detection". From normal to warning to abnormal, it reflects the process of the increasing severity of the problem. Multi-level discrimination can vividly reveal the development trend of the fault, which is convenient for relevant personnel to take targeted prevention and response measures based on this, so as to be prepared for a rainy day and have a targeted approach. At the same time, this method integrates the mining ability of deep learning for complex data, and can discover deep features and laws that are difficult to detect manually. Compared with shallow models, deep models have stronger feature expression and classification capabilities, can process high-dimensional and nonlinear working condition data, and have wider applicability. Since real-time state recognition is performed online, this method places high demands on the timeliness of the algorithm. To this end, the accuracy and speed can be properly balanced in the model design, such as using a lightweight network with a small number of parameters. In addition, the computational efficiency can be improved by means of algorithm optimization, high-performance hardware, etc. In short, this online assessment method can realize the automatic classification and real-time warning of the health status of the combustion chamber, which brings great convenience to risk control and equipment management. Through continuous monitoring and trend analysis, we can grasp the dynamic changes in the health status of equipment, gain insight into the degradation process of equipment, and shift from passive response to active health management. This has important theoretical and practical value for improving the safety, reliability, economy and environmental protection of gas turbines and promoting the advancement of intelligent operation and maintenance technology.

[0074] Step S3300: Output the evaluation result according to the determined health status level, trigger the corresponding early warning information or fault alarm, and update the historical combustion vibration data.

[0075] Specifically, timely feedback of the combustion chamber health assessment results to relevant personnel is a key link for this method to play a practical role. To this end, the assessment system needs to design a user-friendly human-computer interaction interface to display the health status of each combustion chamber in an intuitive and eye-catching way. Commonly used information presentation forms include lists, charts, three-dimensional visualizations, etc., which can be flexibly selected according to the spatial geometric distribution of the assessment object. At the same time, in order to facilitate subsequent traceability and auditing, the results of each assessment must be bound to the necessary information such as the combustion chamber number and assessment timestamp, and then packaged and persistently stored in the assessment record database to form a complete assessment report.

[0076] Example 2

[0077] This embodiment, based on the first embodiment, provides a gas turbine combustion chamber vibration consistency assessment method based on multi-modal fusion, including: Step S1243, calculating a comprehensive metric index sequence according to the silhouette coefficient sequence and the DBI index sequence, and taking the number of clusters corresponding to the maximum value in the comprehensive metric index sequence as the optimal number of clusters N1; Calculating the comprehensive metric index sequence according to the silhouette coefficient sequence and the DBI index sequence comprises: ; in: : Comprehensive metric index, which is the number of clusters The function ranges from The larger it is, the better the clustering effect is.

[0078] : Number of clusters, that is, dividing the data into categories. The value range is ,in is the sample size, Indicates rounding down.

[0079] :When the number of clusters is The silhouette coefficient when .

[0080] :When the number of clusters is DBI index at that time.

[0081] : Weight coefficient of silhouette coefficient.

[0082] : The weight coefficient of the DBI index satisfies The larger it is, the more importance is placed on the cohesion of the category; The larger the value, the more importance is attached to the separation of categories.

[0083] : The power exponent of the silhouette coefficient, used to adjust the influence of the silhouette coefficient. When , the difference of silhouette coefficient is magnified; , the difference in silhouette coefficient is reduced.

[0084] and It needs to be set and adjusted by technicians in this field according to the actual needs. , that is, equal weights.

[0085] It is selected by those skilled in the art based on the distribution characteristics of the data and the sensitivity to the silhouette coefficient. , which is the squared silhouette coefficient.

[0086] : An exponential function with the base e of the natural logarithm as its base; In this formula, the numerator transforms the silhouette coefficient to a power, introducing the parameter , the influence of the silhouette coefficient can be flexibly adjusted. When , the difference of silhouette coefficient is magnified, so that the evaluation of clustering quality pays more attention to the cohesion of categories; when , the difference in silhouette coefficients is reduced, making the evaluation smoother.

[0087] In the numerator, the DBI index is transformed into a negative exponential , mapping it to The range is more consistent with the range of the silhouette coefficient. At the same time, the negative exponential transformation can amplify the difference in the DBI index, making the evaluation of clustering quality pay more attention to the separation of categories.

[0088] In the denominator, introduce the normalization factor , the comprehensive measurement index The value range is limited to Normalization processing makes the comprehensive metric index under different data sets and different parameter settings comparable, which is convenient for evaluating and selecting the optimal number of clusters.

[0089] Through power transformation and negative exponential transformation, the sensitivity of the comprehensive metric index to the difference between the silhouette coefficient and the DBI index is enhanced, making the evaluation of clustering quality more comprehensive and accurate.

[0090] Normalization processing makes the value range of the comprehensive measurement index fixed in There is a unified standard of interpretation and comparison within the framework, which facilitates evaluation and selection under different data sets and different parameter settings. When the optimal number of clusters N1 is taken, the numerator is equal to the denominator; in other cases, . The closer it is to 1, the better the clustering effect; The closer it is to 0, the worse the clustering effect is.

[0091] In summary, this formula enhances the sensitivity of the comprehensive metric index to cluster quality differences and the accuracy of evaluation through power transformation, negative exponential transformation and normalization processing, while providing a unified interpretation and comparison standard. Based on the original silhouette coefficient and DBI index, this formula introduces the normalization idea, which can provide a more reliable and effective quantitative basis for the division of working conditions, and also provides new ideas and methods for the evaluation of other clustering problems.

[0092] Example 3

[0093] This embodiment provides a gas turbine combustion chamber vibration consistency assessment system based on multi-modal fusion on the basis of embodiment 1, such as Figure 8 As shown, including: Feature fusion module: used to obtain a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; the historical combustion vibration data set is segmented to form n1 sub-data sets; for each sub-data set, a preliminary feature vector based on Mahalanobis distance is constructed, and based on the preliminary feature vector, a multi-modal feature vector based on mutual information weighted fusion is constructed; Criteria generation module: used to construct a deep feature extraction model to extract deep features of multimodal feature vectors; perform cluster analysis on deep features to classify the health status level of combustion chamber vibration; calculate the cluster center of each health status level to form a multi-level health status level criterion; the health status level of the combustion chamber vibration includes normal, warning and abnormal; Health level classification module: used to collect the real-time status signal of each combustion chamber under the current operating state, and determine the health status level of the current combustion chamber vibration based on the real-time status signal and multi-level health status level criteria.

[0094] The feature fusion module includes: a data acquisition unit, a data segmentation unit, a preliminary feature construction unit and a multimodal feature construction unit; The data acquisition unit is used to acquire a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; M is a positive integer; The data segmentation unit is used to segment the historical combustion vibration data set to form n1 sub-data sets; The preliminary feature construction unit is used to construct a preliminary feature vector based on Mahalanobis distance for each sub-data set; The multimodal feature construction unit is used to construct a multimodal feature vector based on mutual information weighted fusion according to the preliminary feature vector for each sub-data set.

[0095] The segmentation process of the historical combustion vibration data set to form n1 sub-data sets includes: Step S1210, extracting gas turbine operating parameters corresponding to the historical combustion vibration data set, wherein the operating parameters include load and speed; Step S1220, set the load change threshold δ, perform a traversal analysis on the historical combustion vibration data set in time series, take time t0 as the starting point, if the load change at time t1 relative to the starting time t0 exceeds the load change threshold δ, it is considered that the operating condition has switched, and mark time t1 as the operating condition switching point; then take time t1 as the new starting point, repeat the above process until the end of the historical combustion vibration data set is traversed; Step S1230, dividing the historical combustion vibration data set into n2 initial operating condition segments according to the operating condition switching point; Step S1240, clustering the operating condition parameters in the initial operating condition segment, and dividing the historical combustion vibration data set into n1 subdivided operating condition segments; Step S1250: Generate a corresponding sub-data set for each subdivided operating condition segment to obtain n1 sub-data sets.

[0096] The clustering of the operating parameters in the initial operating condition segment and the division of the historical combustion vibration data set into n1 subdivided operating condition segments include: Step S1241, using the operating parameters in the initial operating condition section as sample features to form a feature matrix X, the dimension of the feature matrix X is m×n, m is the number of samples, and n is the number of features; Step S1242, set the value range of the cluster number K to ,in Indicates rounding down; traverse the cluster number K value and calculate the silhouette coefficient sequence and DBI index sequence; Step S1243, calculating a comprehensive metric index sequence according to the silhouette coefficient sequence and the DBI index sequence, and taking the number of clusters corresponding to the maximum value in the comprehensive metric index sequence as the optimal number of clusters N1; Step S1244, fix the optimal cluster number N1, use the K-Means algorithm to cluster the feature matrix X, obtain n1 subdivided operating condition segments and the cluster center of each subdivided operating condition segment, and form the cluster center vector {C1, C2, ..., C n1}; where N1=n1, C1 is the cluster center of the first subdivided operating condition segment, C2 is the cluster center of the second subdivided operating condition segment, and C n1 It is the cluster center of the n1th subdivided operating condition segment.

[0097] For each sub-data set, constructing a preliminary feature vector based on Mahalanobis distance includes: Step S1310: for each sub-data set, calculate the Mahalanobis distance of each combustion chamber vibration signal relative to the mean of the sub-data set to form a vibration Mahalanobis distance vector ,in, represents the vibration Mahalanobis distance vector of combustion chamber j in the i-th operating condition, i represents the operating condition number, and j represents the combustion chamber number; Step S1320: for each sub-data set, calculate the Mahalanobis distance of each combustion chamber temperature signal relative to the mean of the sub-data set to form a temperature Mahalanobis distance vector ;in, represents the temperature Mahalanobis distance vector of combustion chamber j in the i-th operating condition; Step S1330: The vibration Mahalanobis distance vector and the temperature Mahalanobis distance vector are combined into a preliminary feature vector according to the combustion chamber sequence number. , , Represents the preliminary characteristic vector of combustion chamber j in the i-th operating condition.

[0098] The method of constructing a multimodal feature vector based on mutual information weighted fusion for each sub-data set according to the preliminary feature vector includes: Step S1410: For each combustion chamber, calculate the mutual information matrix between the vibration signal and the temperature signal ; represents the mutual information matrix between the vibration signal and the temperature signal in the i-th operating condition; Step S1420: Mutual information matrix Perform normalization to obtain the fusion weight matrix ; represents the fusion weight matrix of the i-th operating condition; Step S1430: According to the fusion weight matrix Weighted fusion of the i-th operating condition segment The vibration Mahalanobis distance vector of the combustion chamber and temperature Mahalanobis distance vector , construct multimodal fusion features ; Indicates the i-th operating condition segment Multi-modal fusion characteristics of a combustion chamber; Step S1440: Fusion of multi-modal features of each combustion chamber Composition Multi-mode feature vector of each operating condition , , M is the number of combustion chambers, Indicates The multi-modal characteristic vector of the first combustion chamber in each operating section, Indicates The multi-modal characteristic vector of the second combustion chamber in the operating section, Indicates The multi-modal characteristic vector of the Mth combustion chamber in the operating section.

[0099] In the criterion generation module, the deep feature extraction model is constructed to extract the deep features of the multimodal feature vector, including: Step S2110, the deep feature extraction model includes a convolutional neural network and an autoencoder, the multimodal feature vector is input into the convolutional neural network, and a high-dimensional feature map is extracted; Step S2120, inputting the high-dimensional feature map output by the convolutional neural network into the autoencoder to learn global compression low-dimensional features; Step S2130, using the globally compressed low-dimensional features output by the autoencoder as deep features of the multimodal feature vector.

[0100] In the criterion generation module, the calculation of the cluster center of each health status level to form a multi-level health status level criterion includes: Step S2310, respectively calculating the coordinates Cn, Cw, Ca of the cluster centers of the three health status levels of normal, warning, and abnormal; Step S2320, calculating the distance between each sample in the normal state cluster and the normal state cluster center coordinate Cn, and recording the maximum distance as d1; Step S2330, calculating the distance between each sample in the warning state cluster and the center coordinate Cn of the normal state cluster, and recording the minimum distance as d2; Step S2340, set δ1=1 / 2(d1+d2) as the boundary threshold of the normal state; Step S2350, calculating the distance between each sample in the warning state cluster and the coordinate Cw of the center of the warning state cluster, and recording the maximum distance as d3; Step S2360, calculating the distance between each sample in the abnormal state cluster and the center coordinate Cw of the warning state cluster, and recording the minimum distance as d4; Step S2370, set δ2=1 / 2(d3+d4) as the boundary threshold of the warning state; Step S2380, calculate the distance from each sample in the abnormal state cluster to the abnormal state cluster center coordinate Ca, record the maximum distance as d5, and set δ3=2×d5 as the boundary threshold of the abnormal state.

[0101] The health level classification module includes: a real-time feature extraction unit, a real-time health level classification unit and a result output unit; The real-time feature extraction unit is used to collect the real-time state signal of each combustion chamber under the current operating state, wherein the real-time state signal includes a real-time vibration signal and a real-time temperature signal; determine the operating section to which the current operating state belongs, and mark it as the current operating section; calculate the real-time multi-modal feature vector of the real-time state signal; The real-time health level classification unit is used to input the real-time multimodal feature vector into the deep feature extraction model to obtain the real-time deep feature; calculate the Euclidean distance Dn, Dw, Da between the real-time deep feature and the coordinates Cn, Cw, Ca of the cluster centers of the three health status levels of normal, warning and abnormal, and determine the health status level of the current combustion chamber vibration;

[0102] The result output unit is used to output the evaluation result according to the determined health status level, trigger the corresponding early warning information or fault alarm, and update the historical combustion vibration data.

[0103] In the real-time feature extraction unit, determining the operating condition section to which the current operating state belongs includes: Step S3110: Extract real-time operating condition parameters in the current operating state; Step S3120: Perform dimensionless processing on the real-time operating condition parameters to convert the real-time operating condition parameters into a real-time operating condition vector x (c) ; Step S3130: Calculate the Euclidean distances between the real-time operating condition vector and the clustering center vectors {C1, C2,..., C n1} of each fine-grained operating condition segment to obtain a distance vector ; represents the Euclidean distance between the real-time operating condition vector and the clustering center of the first fine-grained operating condition segment, represents the Euclidean distance between the real-time operating condition vector and the clustering center of the second fine-grained operating condition segment, represents the Euclidean distance between the real-time operating condition vector and the clustering center of the n1-th fine-grained operating condition segment; Step S3140: Divide the real-time operating condition vector x (c) into the fine-grained operating condition segment corresponding to the clustering center with the closest distance as the operating condition segment in the current operating state.

[0104] In the real-time feature extraction unit, the calculation of the real-time multi-modal feature vector of the real-time state signal includes: Step S3150: Calculate the Mahalanobis distance between the real-time vibration signal and the mean of the sub-data set belonging to the operating condition segment to form a real-time preliminary feature vector; Step S3160: Construct a real-time multi-modal feature vector based on mutual information weighted fusion according to the fusion weight matrix and the real-time preliminary feature vector.

[0105] In the real-time health level classification unit, the determination of the health state level of the current combustion chamber vibration includes: If Dn ≤ δ1 and Dw ≤ Da, it is determined to be in a normal state; If δ1 < Dn ≤ δ2 and Dw ≤ Da, it is determined to be in a warning state; If δ2 < Dn ≤ δ3, it is determined to be in an abnormal state; If Dn > δ3, it is determined to be in a serious fault state.

[0106] Example 4

[0107] This embodiment discloses an electronic device, which may include one or more processors and one or more memories. Among them, computer-readable code is stored in the memory, and when the computer-readable code is run by one or more processors, it can execute the above-mentioned method for evaluating the vibration consistency of a gas turbine combustion chamber based on multi-modal fusion.

[0108] The method or system according to the embodiment of the present application can also be implemented with the help of the architecture of an electronic device. The electronic device may include a bus, one or more CPUs, a read-only memory (ROM), a random access memory (RAM), a communication port connected to a network, an input / output component, a hard disk, etc. A storage device in the electronic device, such as a ROM or a hard disk, can store the gas turbine combustor vibration consistency assessment method based on multi-modal fusion provided in the present application. A gas turbine combustion chamber vibration consistency assessment method based on multimodal fusion may, for example, include: obtaining a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; segmenting the historical combustion vibration data set to form n1 sub-data sets; for each sub-data set, constructing a preliminary feature vector based on Mahalanobis distance, and based on the preliminary feature vector, constructing a multimodal feature vector based on mutual information weighted fusion; constructing a deep feature extraction model to extract deep features of the multimodal feature vector; performing cluster analysis on the deep features to divide the health status level of the combustion chamber vibration; calculating the cluster center of each health status level to form a multi-level health status level criterion; collecting the real-time status signal of each combustion chamber under the current operating state, and determining the health status level of the current combustion chamber vibration based on the real-time status signal and the multi-level health status level criterion.

[0109] Furthermore, the electronic device may also include a user interface. Of course, the architecture disclosed in the present invention is only exemplary, and when implementing different devices, one or more components in the electronic device disclosed in the present invention may be omitted according to actual needs.

[0110] Example 5

[0111] This embodiment discloses a computer-readable storage medium, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, the gas turbine combustor vibration consistency assessment method based on multi-modal fusion of the embodiment of the present application can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache). Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0112] In addition, according to the implementation of the present application, the process described in the above reference flow chart can be implemented as a computer software program. For example, the present application provides a non-temporary machine-readable storage medium, the non-temporary machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application, for example: obtaining a multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; segmenting the historical combustion vibration data set to form n1 sub-data sets; for each sub-data set, constructing a preliminary feature vector based on Mahalanobis distance, and constructing a multi-modal feature vector based on mutual information weighted fusion according to the preliminary feature vector; constructing a deep feature extraction model to extract deep features of the multi-modal feature vector; performing cluster analysis on the deep features to divide the health status level of the combustion chamber vibration; calculating the cluster center of each health status level to form a multi-level health status level criterion; collecting the real-time status signal of each combustion chamber under the current operating state, and determining the health status level of the current combustion chamber vibration according to the real-time status signal and the multi-level health status level criterion. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are performed.

[0113] The methods, systems, and devices of the present application may be implemented in many ways. For example, the methods, systems, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.

[0114] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0115] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A gas turbine combustor vibration consistency assessment method based on multi-modal fusion, characterized in that: The method comprises: A multi-dimensional historical combustion vibration data set of a gas turbine combustion chamber is obtained; the historical combustion vibration data set includes vibration signals and temperature signals of M combustion chambers; the historical combustion vibration data set is segmented to form n1 sub-data sets; for each sub-data set, a preliminary feature vector based on Mahalanobis distance is constructed, and based on the preliminary feature vector, a multi-modal feature vector based on mutual information weighted fusion is constructed; M and n1 are both positive integers; Construct a deep feature extraction model to extract deep features of multimodal feature vectors; perform cluster analysis on deep features to classify the health status level of combustion chamber vibration; calculate the cluster center of each health status level to form a multi-level health status level criterion; The real-time status signal of each combustion chamber under the current operating state is collected, and the health status level of the current combustion chamber vibration is determined according to the real-time status signal and the multi-level health status level criterion.

2. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 1 is characterized in that: The segmentation process of the historical combustion vibration data set to form n1 sub-data sets includes: Extracting gas turbine operating parameters corresponding to the historical combustion vibration data set, the operating parameters including load and speed; Set the load change threshold δ, perform traversal analysis on the historical combustion vibration data set in time series, take time t0 as the starting point, if the load change at time t1 relative to the starting time t0 exceeds the load change threshold δ, it is considered that the operating condition has switched, and mark time t1 as the operating condition switching point; then take time t1 as the new starting point, repeat the marking process of the operating condition switching point until it ends; According to the operating condition switching point, the historical combustion vibration data set is divided into n2 initial operating condition segments; Cluster the operating parameters in the initial operating condition segment and divide the historical combustion vibration data set into n1 subdivided operating condition segments; A corresponding sub-dataset is generated for each subdivided operating condition segment, and n1 sub-datasets are obtained.

3. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 2 is characterized in that: The clustering of the operating condition parameters in the initial operating condition section includes: The operating parameters in the initial operating condition section are used as sample features to form a feature matrix X. The dimension of the feature matrix X is m×n, where m is the number of samples and n is the number of features. Set the value range of cluster number K to ,in Indicates rounding down; traverse the cluster number K value and calculate the silhouette coefficient sequence and DBI index sequence; The comprehensive metric index sequence is calculated according to the silhouette coefficient sequence and the DBI index sequence, and the cluster number corresponding to the maximum value in the comprehensive metric index sequence is taken as the optimal cluster number N1; The optimal clustering number N1 is fixed, and the K-Means algorithm is used to cluster the feature matrix X to obtain n1 subdivided operating condition segments and the cluster center of each subdivided operating condition segment, forming the cluster center vector of the subdivided operating condition segment; where N1=n1.

4. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 1, characterized in that: The construction of a preliminary feature vector based on Mahalanobis distance includes: Calculate the Mahalanobis distance of the vibration signal of each combustion chamber relative to the mean of the sub-data set to form a vibration Mahalanobis distance vector ,in, represents the vibration Mahalanobis distance vector of combustion chamber j in the i-th operating condition, i represents the operating condition number, and j represents the combustion chamber number; Calculate the Mahalanobis distance of the temperature signal of each combustion chamber relative to the mean of the sub-dataset to form the temperature Mahalanobis distance vector ;in, represents the temperature Mahalanobis distance vector of combustion chamber j in the i-th operating condition; The vibration Mahalanobis distance vector and the temperature Mahalanobis distance vector are combined into a preliminary feature vector according to the combustion chamber sequence. , , Represents the combustion chamber of the i-th operating condition The initial feature vector of .

5. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 4 is characterized in that: The construction of a multimodal feature vector based on mutual information weighted fusion includes: For each combustion chamber, the mutual information matrix of the vibration signal and the temperature signal is calculated ; represents the mutual information matrix between the vibration signal and the temperature signal in the i-th operating condition; Mutual Information Matrix Perform normalization to obtain the fusion weight matrix ; represents the fusion weight matrix of the i-th operating condition; According to the fusion weight matrix Weighted fusion of the i-th operating condition segment The vibration Mahalanobis distance vector of the combustion chamber and temperature Mahalanobis distance vector , construct multimodal fusion features ; Indicates the i-th operating condition segment Multi-modal fusion characteristics of a combustion chamber; The multimodal fusion features of each combustion chamber Composition Multi-mode feature vector of each operating condition .

6. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 1, characterized in that: The deep feature extraction model includes a convolutional neural network and an autoencoder; The deep features of the multimodal feature vector extraction include: The multimodal feature vector is input into the convolutional neural network to extract the high-dimensional feature map; the high-dimensional feature map output by the convolutional neural network is input into the autoencoder to learn the global compression low-dimensional features; The globally compressed low-dimensional features output by the autoencoder are used as the deep features of the multimodal feature vector.

7. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 1, characterized in that: The health status levels of the combustion chamber vibration include normal, warning and abnormal; Calculating the cluster centers of each health state level to form a multi-level health state level criterion includes: Calculating the coordinates Cn, Cw, and Ca of the cluster centers of the three health state levels of normal, warning, and abnormal respectively; Calculating the distance from each sample in the normal state cluster to the coordinate Cn of the normal state cluster center, and recording the maximum distance as d1; Calculating the distance from each sample in the warning state cluster to the coordinate Cn of the normal state cluster center, and recording the minimum distance as d2; Let δ1 = 1 / 2(d1 + d2), which is used as the boundary threshold of the normal state; Calculating the distance from each sample in the warning state cluster to the coordinate Cw of the warning state cluster center, and recording the maximum distance as d3; Calculating the distance from each sample in the abnormal state cluster to the coordinate Cw of the warning state cluster center, and recording the minimum distance as d4; Let δ2 = 1 / 2(d3 + d4), which is used as the boundary threshold of the warning state; Calculating the distance from each sample in the abnormal state cluster to the coordinate Ca of the abnormal state cluster center, and recording the maximum distance as d5, and let δ3 = 2 × d5, which is used as the boundary threshold of the abnormal state.

8. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 7 is characterized in that: The real-time state signal includes a real-time vibration signal and a real-time temperature signal; Determining the health state level of the current combustion chamber vibration according to the real-time state signal and the multi-level health state level criterion includes: Determining the operating condition section to which the current operating state belongs, and marking it as the current operating condition section; calculating the real-time multi-modal feature vector of the real-time state signal; Inputting the real-time multi-modal feature vector into the deep feature extraction model to obtain the real-time deep feature; calculating the Euclidean distances Dn, Dw, and Da between the real-time deep feature and the coordinates Cn, Cw, and Ca of the cluster centers of the three health state levels of normal, warning, and abnormal, and determining the health state level of the current combustion chamber vibration.

9. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 8, characterized in that: Determining the operating condition section to which the current operating state belongs includes: Extract the real-time operating parameters under the current operating state; perform dimensionless processing on the real-time operating parameters and convert them into the real-time operating vector x (c) ; Calculate the Euclidean distance between the real-time operating condition vector and the cluster center vector of the subdivided operating condition segment; (c) Divide into the subdivided operating condition segment corresponding to the nearest cluster center as the operating condition segment of the current operating state; Calculating the real-time multi-modal feature vector of the real-time state signal includes: Calculate the Mahalanobis distance of the real-time vibration signal relative to the mean of the sub-data set of the operating section to form a real-time preliminary feature vector; according to the fusion weight matrix and real-time preliminary feature vectors, and construct real-time multimodal feature vectors based on mutual information weighted fusion.

10. The gas turbine combustor vibration consistency assessment method based on multi-modal fusion according to claim 8, characterized in that: Determining the health state level of the current combustion chamber vibration includes: Determining the health state level of the current combustion chamber vibration includes: If Dn ≤ δ1 and Dw ≤ Da, it is determined to be in the normal state; If δ1 < Dn ≤ δ2 and Dw ≤ Da, it is determined to be in the warning state; If δ2 < Dn ≤ δ3, it is determined to be in the abnormal state.

Images