Gas turbine fault identification method, device, equipment and medium

Abstract

The present disclosure proposes a method, device, equipment and medium for fault identification of a gas turbine, which relates to the field of data processing technology. The method includes: monitoring the gas turbine to obtain monitoring values ​​of multiple dimensions; generating a feature vector related to a specified target fault type according to the monitoring values ​​of multiple dimensions; querying a target fault identification matrix matching the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the feature vector; determining the fault probabilities of multiple sub-fault types according to the index values ​​of the multiple sub-fault types; and determining the target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types. Thus, the probability of each sub-fault type under the target fault type occurring in the gas turbine is effectively determined, and then the fault type to which the gas turbine belongs can be effectively identified based on the probability of each sub-fault type.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a method, device, equipment and medium for identifying a fault of a gas turbine. Background Art

[0002] Gas turbines are an important type of internal combustion power machinery that uses continuously flowing gas as a working fluid to drive the impeller to rotate at high speed, converting the energy of the fuel into useful work. As an advanced power device, gas turbines have the characteristics of fast start-up, low noise, small size, and high power. However, gas turbines work in harsh environments with high temperature, high pressure, high speed, and high load for a long time, and are prone to failures, with many types of failures and a high probability of coupling.

[0003] Therefore, it is very important to identify the fault type of the gas turbine so that relevant personnel can promptly perform corresponding repairs and maintenance on the gas turbine according to the fault type of the gas turbine to ensure the normal operation of the gas unit and the safety of personnel. Summary of the invention

[0004] The present disclosure aims to solve one of the technical problems in the related art at least to some extent.

[0005] The present disclosure proposes a gas turbine fault identification method, device, equipment and medium, which can effectively determine the probability of each sub-fault type under the target fault type of the gas turbine based on the monitoring values ​​of multiple dimensions obtained by monitoring the gas turbine, and then can effectively identify the fault type of the gas turbine based on the probability of each sub-fault type, thereby improving the accuracy and reliability of the fault identification result.

[0006] The first embodiment of the present disclosure provides a method for identifying a fault of a gas turbine, comprising:

[0007] Monitoring the gas turbine to obtain monitoring values ​​in multiple dimensions;

[0008] Generating a feature vector related to a specified target fault type according to the monitoring values ​​of the multiple dimensions;

[0009] Querying a target fault identification matrix matching the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector;

[0010] Determining the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types;

[0011] According to the failure probabilities of the multiple sub-fault types, a target sub-fault type to which the gas turbine belongs is determined from the multiple sub-fault types.

[0012] The fault identification method of the gas turbine of the embodiment of the present disclosure monitors the gas turbine to obtain monitoring values ​​of multiple dimensions; generates a feature vector related to the specified target fault type according to the monitoring values ​​of the multiple dimensions; queries the target fault identification matrix matching the target fault type, and generates index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the feature vector; determines the fault probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types; and determines the target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types. Thus, the probability of each sub-fault type under the target fault type of the gas turbine can be effectively determined according to the monitoring values ​​of multiple dimensions obtained by monitoring the gas turbine, and then the fault type to which the gas turbine belongs can be effectively identified based on the probability of each sub-fault type, thereby improving the accuracy and reliability of the fault identification result.

[0013] A second aspect of the present disclosure provides a gas turbine fault identification device, comprising:

[0014] A monitoring module, used for monitoring the gas turbine to obtain monitoring values ​​in multiple dimensions;

[0015] A generating module, used for generating a feature vector related to a specified target fault type according to the monitoring values ​​of the multiple dimensions;

[0016] A processing module, used for querying a target fault identification matrix matching the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector;

[0017] A first determination module, configured to determine the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types;

[0018] The second determination module is used to determine a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types.

[0019] The fault identification device of the gas turbine of the disclosed embodiment monitors the gas turbine to obtain monitoring values ​​of multiple dimensions; generates a characteristic vector related to the specified target fault type according to the monitoring values ​​of multiple dimensions; queries the target fault identification matrix matching the target fault type, and generates index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector; determines the fault probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types; and determines the target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types. Thus, the probability of each sub-fault type under the target fault type occurring in the gas turbine can be effectively determined according to the monitoring values ​​of multiple dimensions obtained by monitoring the gas turbine, and then the fault type to which the gas turbine belongs can be effectively identified based on the probability of each sub-fault type, thereby improving the accuracy and reliability of the fault identification result.

[0020] The third aspect embodiment of the present disclosure proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the gas turbine fault identification method proposed in the first aspect embodiment of the present disclosure is implemented.

[0021] The fourth aspect embodiment of the present disclosure proposes a non-temporary computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the gas turbine fault identification method proposed in the first aspect embodiment of the present disclosure.

[0022] The fifth aspect embodiment of the present disclosure proposes a computer program product. When the instructions in the computer program product are executed by a processor, the fault identification method of the gas turbine proposed in the first aspect embodiment of the present disclosure is executed.

[0023] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0025] Figure 1 A schematic flow chart of a method for identifying a fault of a gas turbine provided in the first embodiment of the present disclosure;

[0026] Figure 2 A schematic flow chart of a method for identifying a fault of a gas turbine provided in the second embodiment of the present disclosure;

[0027] Figure 3A schematic flow chart of a method for identifying a fault of a gas turbine provided in Embodiment 3 of the present disclosure;

[0028] Figure 4 is a flow chart of fault identification of a gas turbine in an embodiment of the present disclosure;

[0029] Figure 5 This is a schematic diagram of the structure of a gas turbine fault identification device provided in Embodiment 4 of the present disclosure. DETAILED DESCRIPTION

[0030] Embodiments of the present disclosure are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0031] Gas turbines are important power machinery equipment. However, they work in harsh environments with high temperature, high pressure, high speed and high load for a long time, which makes them prone to failures. There are many types of failures and a high coupling probability. At present, there are two main types of failures in gas turbines in engineering applications. One is the degradation of gas turbine performance, which is a gas path failure. The other is excessive vibration, which is a structural strength failure. These two types of failures often appear in a coupled form, which increases the difficulty of fault identification.

[0032] Traditional gas turbine fault identification methods mainly include the following three methods: fault identification method based on mechanism model, fault identification method based on data drive and fault identification method based on experience knowledge. The basic principles of the above three gas turbine fault identification methods are the same, that is, to extract fault characteristic indicators with obvious differentiation from the monitoring data of the gas turbine, and then combine the fault judgment criteria corresponding to each fault type to finally determine the fault type of the gas turbine. For example, assuming that the fault characteristic indicators extracted from the monitoring data of the gas turbine have a high similarity with the fault judgment criteria corresponding to fault type A, it is determined that the fault type of the gas turbine is fault type A.

[0033] However, the monitoring data of gas turbines includes process quantities and state quantities, which are massive big data. Traditional fault identification methods need to rely on manual diagnosis, and it is difficult to reach a diagnostic conclusion in a timely manner.

[0034] In related technologies, the automatic fault identification method is based on pattern recognition theory. The essence of pattern recognition theory is to automatically identify fault types by dimensionality reduction. Common pattern recognition methods include distance measurement algorithms and similarity measurement algorithms:

[0035] (I) Distance measurement algorithm:

[0036] Distance measurement algorithms can be used to calculate the distance between two data samples in multidimensional space. Common distance measurement algorithms include Euclidean distance, Manhattan distance, Chebyshev distance, and Minkowski distance. The principles of various distance measurement algorithms are:

[0037] 1) Euclidean Distance (ED)

[0038] The ED algorithm is the most widely used distance measurement algorithm, which can be used to calculate the absolute spatial distance between two samples. When using the ED algorithm, the sample data can be standardized, so that the ED between samples can be calculated based on the standardized data. Assume that sample A is (x1, x2, …, x n ), sample B is (y1,y2,…,y n ), then the ED between sample A and sample B can be calculated according to the following formula:

[0039]

[0040] Among them, sample B can be a fault criterion corresponding to a certain fault type, and sample A can be a vector obtained by reducing the monitoring values ​​of multiple dimensions in the monitoring data of the gas turbine to n dimensions. For example, when the gas turbine does not include a power turbine, if it is necessary to identify whether the gas turbine has a specific sub-fault type fault under a gas path fault, then n=4, x1 can be the compressor flow, x2 can be the compressor efficiency, x3 can be the turbine flow, and x4 can be the turbine efficiency.

[0041] When the ED between sample A and sample B is smaller, it indicates that the similarity between sample A and sample B is higher, and the probability of the gas turbine having a fault of the fault type corresponding to sample B is greater.

[0042] 2) Manhattan Distance (MAD)

[0043] MAD can be used to calculate the absolute axial distances between two samples in different dimensions of multidimensional space and sum the axial distances. Assume that sample A is (x1, x2, …, x n ), sample B is (y1,y2,…,y n ), then the MAD between sample A and sample B can be calculated according to the following formula:

[0044]

[0045] When the MAD between sample A and sample B is smaller, it indicates that the similarity between sample A and sample B is higher, and the probability of the gas turbine having a fault of the fault type corresponding to sample B is greater.

[0046] 3) Minkowski Distance (MID)

[0047] MID is an improvement of ED, and MID is obtained by summarizing multiple distance measurement formulas. Assume that sample A is (x1, x2, …, x n ), sample B is (y1,y2,…,y n ), then the MID between sample A and sample B can be calculated according to the following formula:

[0048]

[0049] Where p is the type of distance metric. When p=1, MID corresponds to Manhattan distance; when p=2, MID corresponds to Euclidean distance.

[0050] When the MID between sample A and sample B is smaller, it indicates that the similarity between sample A and sample B is higher, and the probability of the gas turbine having a fault of the fault type corresponding to sample B is greater.

[0051] 4) Chebyshev Distance (CD)

[0052] CD is a further generalization of MID. Assume that sample A is (x1, x2, …, x n ), sample B is (y1,y2,…,y n ), then the CD between sample A and sample B can be calculated according to the following formula:

[0053]

[0054] It should be noted that the greater the distance between two samples, the smaller the similarity between the two samples. On the contrary, the smaller the distance between two samples, the greater the similarity between the two samples. Moreover, when using a distance measurement algorithm to determine the distance between two samples, any of the above algorithms can be used.

[0055] (II) Similarity measurement algorithm:

[0056] Similarity measurement algorithms can be used to calculate the similarity between two samples. The larger the similarity measurement value, the greater the pattern similarity. Similarity measurement algorithms mainly include correlation coefficient and cosine similarity. The principles of each similarity algorithm are:

[0057] 1) Cosine Similarity (CS)

[0058] CS is used to calculate the similarity between two samples. The cosine value can be calculated based on the angle between the two samples. The cosine value is the CS value. Among them, CS is sensitive to the spatial direction and insensitive to the vector modulus. Assume that sample A is (x1, x2, …, x n ), sample B is (y1,y2,…,y n ), then the CS between sample A and sample B can be calculated according to the following formula:

[0059]

[0060] The value range of CS is between -1 and 1. When CS=1, the two samples are exactly the same; when CS=-1, the two samples are completely opposite; when CS=0, the two samples are independent of each other.

[0061] 2) Pearson Correlation Coefficient (PCC)

[0062] By calculating the covariance and standard deviation of two samples, and then dividing the covariance and standard deviation of the two samples, we can get the PCC value, which can be used to characterize the linear correlation between samples and is independent of the position and scale of the samples in three-dimensional space. Assume that sample A is (x1, x2, …, x n ), sample B is (y1,y2,…,y n ), then the PCC between sample A and sample B can be calculated according to the following formula:

[0063]

[0064] Among them, the value range of PCC is between -1 and 1. When the PCC value approaches 1, the two samples tend to be the same; when the PCC value approaches -1, the two samples tend to be completely opposite; when the PCC value is 0, the two samples are unrelated.

[0065] It should be noted that when a similarity measurement algorithm is used to determine the similarity value between two samples, any one of the above methods may be used.

[0066] As an example, a typical gas path failure of a gas turbine is used as an example. The gas path failure may be accompanied by a change in the flow structure of the component, and correspondingly, the performance parameters of the component may change. Among them, the performance parameters related to the gas path failure may include flow parameters and efficiency parameters. Based on a large amount of experimental research data, the impact of typical gas path failures on performance parameters can be shown in Table 1:

[0067] Table 1 Influence of typical gas circuit failures on performance parameters

[0068] Serial number Sub-fault type Criteria 1 Compressor blade fouling Compressor flow rate is reduced by 7% and compressor efficiency is reduced by 2% 2 Compressor blade wear Compressor efficiency degraded by 2% 3 Compressor blade mechanical damage Compressor efficiency degraded by 5% 4 Increased compressor tip clearance Compressor flow rate downgraded by 4% 5 Turbine blade fouling Turbine flow rate is reduced by 6% and turbine efficiency is reduced by 2% 6 Turbine blade wear Turbine flow rate increased by 6%, turbine efficiency decreased by 2% 7 Mechanical damage to turbine blades Turbine efficiency degraded by 5% 8 Thermal corrosion of turbine blades Turbine flow upgrade 6%

[0069] Among them, compressor blade fouling (i.e., fault 1), compressor blade wear (i.e., fault 2), compressor blade mechanical damage (i.e., fault 3), compressor blade tip clearance increase (i.e., fault 4), turbine blade fouling (i.e., fault 5), turbine blade wear (i.e., fault 6), turbine blade mechanical damage (i.e., fault 7), and turbine blade thermal corrosion (i.e., fault 8) are sub-fault types under the gas path fault.

[0070] Among them, the pattern recognition method can be used to automatically identify the eight sub-fault types under the above-mentioned typical gas path faults.

[0071] As an example, the inventors use the ED method in the distance measurement algorithm to calculate the pattern similarity between the above 8 sub-fault types. The calculation results can be shown in Table 2:

[0072] Table 2 Calculation of pattern similarity between sub-fault types based on ED method

[0073] Sub-fault type Fault 1 Fault 2 Fault 3 Fault 4 Fault 5 Fault 6 Fault 7 Fault 8 Fault 1 0.000 3.021 3.021 0.643 3.678 2.905 3.552 2.718 Fault 2 3.021 0.000 0.000 3.464 3.563 2.967 3.464 2.828 Fault 3 3.021 0.000 0.000 3.464 3.563 2.967 3.464 2.828 Fault 4 0.643 3.464 3.464 0.000 3.563 2.967 3.464 2.828 Fault 5 3.678 3.563 3.563 3.563 0.000 4.249 2.941 4.409 Fault 6 2.905 2.967 2.967 2.967 4.249 0.000 2.285 0.662 Fault 7 3.552 3.464 3.464 3.464 2.941 2.285 0.000 2.828 Fault 8 2.718 2.828 2.828 2.828 4.409 0.662 2.828 0.000

[0074] Among them, the ED between fault i (i is a positive integer not greater than 8) and fault j (j is a positive integer not greater than 8) in Table 2 refers to the ED value between the fault criterion corresponding to fault i and the fault criterion corresponding to fault j.

[0075] For example, the ED algorithm is used to identify fault 1 and fault 4 under gas turbine gas path fault. When fault 1 occurs in the gas turbine, that is, fouling of the compressor blades, the performance parameters related to fault 1 include: compressor flow (e.g., marked as x1), compressor efficiency (e.g., marked as x2), turbine flow (e.g., marked as x3), turbine efficiency (e.g., marked as x4), so that according to the performance parameters of the components related to the gas path fault (i.e., compressor flow, compressor efficiency, turbine flow, turbine efficiency), the fault criterion corresponding to fault 1 is (x1, x2, x3, x4); when fault 4 occurs in the gas turbine, that is, the compressor blade tip When the gap increases, the performance parameters related to fault 4 include: compressor flow (for example, marked as y1), compressor efficiency (for example, marked as y2), turbine flow (for example, marked as y3), and turbine efficiency (for example, marked as y4). Therefore, according to the performance parameters related to the gas path fault (i.e., compressor flow, compressor efficiency, turbine flow, and turbine efficiency), the corresponding fault criterion for fault 4 is generated as (y1, y2, y3, y4); the distance between the fault criterion of fault 1 (x1, x2, x3, x4) and the fault criterion of fault 4 (y1, y2, y3, y4) calculated by the ED algorithm is 0.643 in Table 2.

[0076] It can be understood that, the greater the distance between the fault criterion of fault 1 and the fault criterion of fault 4, the smaller the similarity between the fault criterion of fault 1 and the fault criterion of fault 4, and the easier it is to distinguish fault 1 from fault 4 in the actual gas turbine gas path fault identification; and when the smaller the distance between the fault criterion of fault 1 and the fault criterion of fault 4, the greater the similarity between the fault criterion of fault 1 and the fault criterion of fault 4, and the easier it is to confuse the sub-fault type of the gas path fault of the gas turbine in the actual gas turbine gas path fault identification, that is, when the fault type of the gas turbine is fault 1, the fault type of the gas turbine may be mistakenly identified as fault 4, or, when the fault type of the gas turbine is fault 4, the fault type of the gas turbine may be mistakenly identified as fault 1.

[0077] However, according to the calculation results in Table 2, since the ED values ​​between the fault criteria of fault 1 and fault 4, and the ED values ​​between the fault criteria of fault 6 and fault 8 are both less than 0.7, and the ED value between the fault criteria of fault 2 and fault 3 is close to 0, it is difficult to distinguish fault 1 from fault 4, fault 6 from fault 8, and fault 2 from fault 3, and the mode isolation effect is not good.

[0078] As another example, the inventors use the PCC method in the distance measurement algorithm to calculate the pattern similarity between the above 8 sub-fault types. The calculation results can be shown in Table 3:

[0079] Table 3. Pattern similarity between sub-fault types calculated based on PCC method

[0080]

[0081]

[0082] Among them, the PCC between fault i (i is a positive integer not greater than 8) and fault j (j is a positive integer not greater than 8) in Table 3 refers to the PCC value between the fault criterion corresponding to fault i and the fault criterion corresponding to fault j.

[0083] For example, the PCC algorithm is used to calculate the similarity of the fault criteria (x1, x2, x3, x4) of fault 1 and the fault criteria (y1, y2, y3, y4) of fault 4, and the PCC value between the fault criteria of fault 1 and the fault criteria of fault 4 can be 0.959 as shown in Table 3.

[0084] It can be understood that the greater the similarity value between the fault criterion of fault 1 and the fault criterion of fault 4 (i.e., the greater the PCC value), the easier it is to confuse the sub-fault type of the gas path fault of the gas turbine in the actual gas turbine gas path fault identification, that is, when the fault type of the gas turbine is fault 1, the fault type of the gas turbine may be misidentified as fault 4, or, when the fault type of the gas turbine is fault 4, the fault type of the gas turbine may be misidentified as fault 1; and when the similarity value between the fault criterion of fault 1 and the fault criterion of fault 4 is smaller, it is easier to distinguish fault 1 from fault 4 in the actual gas turbine gas path fault identification.

[0085] However, according to the calculation results in Table 3, since the PCC values ​​between the fault criteria of fault 1 and fault 4, and between the fault criteria of fault 6 and fault 8 are both greater than 0.95, and the PCC value between the fault criteria of fault 2 and fault 3 is close to 1, it is difficult to distinguish between fault 1 and fault 4, fault 6 and fault 8, and fault 2 and fault 3, and the mode isolation effect is not good.

[0086] In summary, when using the pattern recognition method for gas turbine fault identification, the pattern isolation effect is poor. For example, the distance value between the fault criteria of two different sub-faults under the gas path fault is small and / or the similarity value is large, which makes it difficult to clearly distinguish the sub-faults under the gas path fault. The reasons for the poor pattern isolation effect are:

[0087] (1) The pattern similarity based on the distance measurement algorithm only evaluates the absolute distance between samples, that is, it only considers the modulus of the sample vector but not the direction of the sample vector;

[0088] (2) The pattern similarity based on the similarity measurement algorithm only evaluates the directional characteristics of the sample vector without considering the modulus of the sample vector.

[0089] In view of the above problems, the present disclosure proposes a gas turbine fault identification method, device, equipment and medium.

[0090] The following describes a gas turbine fault identification method, apparatus, device, and medium according to an embodiment of the present disclosure with reference to the accompanying drawings.

[0091] Figure 1 A schematic flow chart of a gas turbine fault identification method provided in Embodiment 1 of the present disclosure.

[0092] The embodiment of the present disclosure takes the gas turbine fault identification method being configured in a gas turbine fault identification device as an example. The gas turbine fault identification device can be applied to any electronic device so that the electronic device can perform the gas turbine fault identification function.

[0093] Among them, the electronic device can be any device with computing capabilities, such as a computer, a mobile terminal, a server, etc. The mobile terminal can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, and other hardware devices with various operating systems, touch screens and / or display screens.

[0094] like Figure 1 As shown, the gas turbine fault identification method may include the following steps:

[0095] Step 101: monitor the gas turbine to obtain monitoring values ​​in multiple dimensions.

[0096] In the embodiments of the present disclosure, monitoring values ​​of multiple dimensions may include, for example, the compressor inlet temperature, compressor outlet temperature, compressor inlet pressure, compressor outlet pressure, compressor speed, compressor inlet air flow, compressor outlet air flow, oil temperature at the inlet of the lubricating oil system, oil temperature at the outlet of the lubricating oil system, bearing vibration frequency, bearing vibration speed, bearing vibration acceleration, turbine inlet pressure, turbine outlet pressure, etc. of the gas turbine, and the present disclosure does not impose any restrictions on this.

[0097] In the embodiments of the present disclosure, the gas turbine may be monitored. For example, a plurality of sensors may be used to monitor the gas turbine in real time to obtain monitoring values ​​in multiple dimensions.

[0098] Step 102: Generate a feature vector related to a specified target fault type according to monitoring values ​​of multiple dimensions.

[0099] In the disclosed embodiment, the target fault type may be a gas path fault and / or a structural strength fault.

[0100] In the embodiment of the present disclosure, a feature vector related to a specified target fault type may be generated according to monitoring values ​​of multiple dimensions.

[0101] As an example, taking the specified target fault type as a gas circuit fault as an example, a feature vector related to the gas circuit fault can be generated according to monitoring values ​​of multiple dimensions.

[0102] As another example, taking the designated target fault type as a structural strength fault, a feature vector related to the structural strength fault can be generated according to monitoring values ​​in multiple dimensions.

[0103] As another example, taking the designated target fault types as gas path fault and structural strength fault, feature vectors related to the gas path fault and structural strength fault may be generated according to detection values ​​of multiple dimensions.

[0104] Step 103 , querying a target fault identification matrix that matches the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector.

[0105] In the embodiment of the present disclosure, the target fault identification matrix may be preset, and the target fault identification matrix matches the target fault type.

[0106] It should be noted that the target fault identification matrix can be set based on manual experience.

[0107] As an example, taking the target fault type as a gas path fault, when the gas turbine does not include a power turbine, the target fault identification matrix can be determined according to Table 1. For example, the target fault identification matrix can be:

[0108]

[0109] In an embodiment of the present disclosure, a target fault type may include multiple sub-fault types. For example, taking the target fault type as an air path fault, the air path fault may include sub-fault types such as compressor blade fouling, compressor blade wear, compressor blade mechanical damage, increased compressor blade tip clearance, turbine blade fouling, turbine blade wear, turbine blade mechanical damage, and turbine blade thermal corrosion. The present disclosure does not impose any restrictions on this.

[0110] It should be noted that the above examples of various sub-fault types under gas circuit faults are merely illustrative. In actual applications, the sub-fault types under gas circuit faults may be some of the above sub-fault types, or may include other sub-fault types, and the present disclosure does not impose any restrictions on this.

[0111] In the embodiment of the present disclosure, the fault identification matrix corresponding to each fault type can be preset, and the corresponding relationship between the fault type and the fault identification matrix can be stored, so that in the present disclosure, in the above corresponding relationship, the target fault identification matrix matching the target fault type can be queried. After the target fault identification matrix matching the target fault type is queried, the index values ​​of multiple sub-fault types under the target fault type can be generated according to the target fault identification matrix and the characteristic vector related to the target fault type.

[0112] Step 104 : determining the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types.

[0113] In the embodiment of the present disclosure, the failure probabilities of the multiple sub-fault types may be determined according to the index values ​​of the multiple sub-fault types.

[0114] Step 105 : determining a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types.

[0115] In the embodiment of the present disclosure, the target sub-fault type is the fault type to which the gas turbine belongs. In the present disclosure, the target sub-fault type to which the gas turbine belongs can be determined from multiple sub-fault types according to the fault probabilities of the multiple sub-fault types.

[0116] In a possible implementation manner of the embodiment of the present disclosure, a sub-fault type whose fault probability is greater than a set probability threshold may be determined as a target sub-fault type to which the gas turbine belongs.

[0117] In the embodiment of the present disclosure, the probability threshold may be pre-set, for example, the probability threshold is 50%, 60%, etc., and the present disclosure does not impose any limitation on this.

[0118] It should be noted that the probability threshold may be set based on manual experience, or may be dynamically adjusted based on actual application scenarios and application requirements, and the present disclosure does not impose any restrictions on this.

[0119] In another possible implementation of the embodiment of the present disclosure, the sub-fault type corresponding to the maximum fault probability among the fault probabilities may be determined as the target sub-fault type to which the gas turbine belongs.

[0120] The fault identification method of the gas turbine of the embodiment of the present disclosure monitors the gas turbine to obtain monitoring values ​​of multiple dimensions; generates a feature vector related to the specified target fault type according to the monitoring values ​​of the multiple dimensions; queries the target fault identification matrix matching the target fault type, and generates index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the feature vector; determines the fault probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types; and determines the target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types. Thus, the probability of each sub-fault type under the target fault type of the gas turbine can be effectively determined according to the monitoring values ​​of multiple dimensions obtained by monitoring the gas turbine, and then the fault type to which the gas turbine belongs can be effectively identified based on the probability of each sub-fault type, thereby improving the accuracy and reliability of the fault identification result.

[0121] In order to clearly illustrate how, in the above embodiments of the present disclosure, index values ​​of multiple sub-fault types under a target fault type are generated based on a target fault identification matrix and a characteristic vector related to the target fault type, the present disclosure also proposes a gas turbine fault identification method.

[0122] Figure 2 A flow chart of a gas turbine fault identification method provided in the second embodiment of the present disclosure.

[0123] like Figure 2 As shown, the gas turbine fault identification method may include the following steps:

[0124] Step 201: monitor the gas turbine to obtain monitoring values ​​in multiple dimensions.

[0125] Step 202: Generate a feature vector related to a specified target fault type according to monitoring values ​​of multiple dimensions, wherein the feature vector includes n elements.

[0126] The execution process of steps 201 to 202 may refer to the execution process of any embodiment of the present disclosure, and will not be described in detail here.

[0127] In the embodiment of the present disclosure, the number of elements contained in the feature vector related to the target fault type is n. For example, taking the target fault type as an air path fault and the gas turbine does not include a power turbine as an example, the feature vector may be (a1, a2, a3, a4), wherein a1 is the compressor flow, a2 is the compressor efficiency, a3 is the turbine flow, and a4 is the turbine efficiency. For another example, taking the target fault type as an air path fault and the gas turbine includes a power turbine as an example, the feature vector may be (a1, a2, a3, a4, a5, a6), wherein a5 is the power turbine flow, and a6 is the power turbine efficiency.

[0128] Step 203 , querying a target fault identification matrix that matches the target fault type, and generating a fault vector corresponding to the i-th sub-fault type under the target fault type according to the i-th row elements in the target fault identification matrix.

[0129] The number of sub-fault types under the target fault type is m, i is a positive integer not greater than m, and the target fault identification matrix is ​​an m*n matrix. For example, taking the target fault type as a gas path fault and the gas turbine not including a power turbine as an example, as shown in step 103, the target fault identification matrix may be an 8*4 matrix. For another example, taking the target fault type as a gas path fault and the gas turbine including a power turbine as an example, the target fault identification matrix may be an 8*6 matrix.

[0130] In the embodiment of the present disclosure, after querying the target fault identification matrix matching the target fault type, a fault vector corresponding to the i-th seed fault type can be generated according to the i-th row elements in the target fault identification matrix; wherein i is a positive integer not greater than m.

[0131] For example, the target fault identification matrix is ​​labeled as Then, according to the i-th row element in the target fault identification matrix, the fault vector corresponding to the i-th seed fault type can be (A i1 ,A i2 ,…,A in ).

[0132] Step 204 : Determine the similarity between the fault vector and the feature vector of the i-th fault type.

[0133] In the disclosed embodiment, the similarity between the fault vector and the feature vector of the i-th seed fault type may be determined. For example, a cosine similarity (CS) algorithm, a Pearson Correlation Coefficient (PCC) algorithm, etc. may be used to determine the similarity between the fault vector and the feature vector of the i-th seed fault type.

[0134] As an example, the similarity between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by using the cosine similarity algorithm, and the fault vector of the i-th seed fault type is marked as (A i1 ,A i2 ,…,A in ), the feature vector associated with the target fault type is (B1,B2,…,B n ), then the similarity S between the fault vector of the i-th seed fault type and the above feature vector i It can be determined according to the following formula:

[0135]

[0136] As another example, the similarity between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by using the Pearson correlation coefficient algorithm. The fault vector of the i-th seed fault type is marked as (A i1 ,A i2 ,…,A in ), the eigenvector is (B1,B2,…,B n ), then the similarity S between the fault vector and the feature vector of the i-th seed fault type i It can be determined according to the following formula:

[0137]

[0138] in,

[0139]

[0140] Step 205: Determine the index value of the i-th seed fault type according to the similarity.

[0141] In the embodiment of the present disclosure, the index value of the i-th fault type may be determined according to the similarity between the fault vector and the feature vector of the i-th fault type. For example, the index value may be in a positive relationship with the similarity.

[0142] In order to clearly explain how the index value of the i-th seed fault type is determined based on the above-mentioned similarity in the present disclosure, in a possible implementation method of an embodiment of the present disclosure, the distance between the fault vector of the i-th seed fault type and the above-mentioned characteristic vector can be determined, and the first coefficient can be determined based on the sum of the above-mentioned similarity and the above-mentioned distance, and thus the index value of the i-th seed fault type can be determined based on the ratio of the above-mentioned similarity to the first coefficient.

[0143] In the disclosed embodiment, the distance between the fault vector of the i-th seed fault type and the feature vector associated with the target fault type can be determined. For example, the Euclidean distance (ED) algorithm, the Manhattan distance (MAD) algorithm, the Minkowski distance (MID) algorithm, the Chebyshev distance (CD) algorithm, etc. can be used to determine the distance between the fault vector of the i-th seed fault type and the feature vector associated with the target fault type.

[0144] As an example, the distance between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by using the Euclidean distance algorithm, and the fault vector of the i-th seed fault type is marked as (A i1 ,A i2 ,…,A in ), the feature vector associated with the target fault type is (B1,B2,…,B n ), then the distance L between the fault vector of the i-th fault type and the above feature vector i It can be determined according to the following formula:

[0145]

[0146] As another example, the distance between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by using the Manhattan distance algorithm. The fault vector of the i-th seed fault type is marked as (A i1 ,A i2 ,…,A in ), the feature vector associated with the target fault type is (B1,B2,…,B n ), then the distance L between the fault vector of the i-th fault type and the above feature vector i It can be determined according to the following formula:

[0147]

[0148] As another example, the distance between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by using the Chebyshev distance algorithm. The fault vector of the i-th seed fault type is marked as (A i1 ,A i2 ,…,A in ), the feature vector associated with the target fault type is (B1,B2,…,B n ), then the distance L between the fault vector of the i-th fault type and the above feature vector i It can be determined according to the following formula:

[0149]

[0150] In the embodiment of the present disclosure, after determining the distance between the fault vector of the i-th fault type and the above-mentioned characteristic vector, the first coefficient can be determined according to the similarity between the fault vector of the i-th fault type and the above-mentioned characteristic vector and the sum of the above-mentioned distances.

[0151] For example, the similarity between the fault vector marking the i-th seed fault type and the above feature vector is S i , the distance between the fault vector of the i-th fault type and the above feature vector is L i , then the first coefficient K i Can be:

[0152] K i =S i +L i ; (14)

[0153] In the embodiment of the present disclosure, after the first coefficient is determined, the index value of the i-th seed fault type may be determined according to the ratio of the similarity between the fault vector of the i-th seed fault type and the above-mentioned feature vector and the first coefficient.

[0154] For example, the similarity between the fault vector marking the i-th seed fault type and the above feature vector is Si , the first coefficient is K i , then the index value CRI of the i-th seed fault type is i Can be:

[0155]

[0156] It should be noted that, for each sub-fault type under the target fault type, the method of steps 203 to 205 may be used to determine its corresponding index value, which will not be described in detail here.

[0157] Step 206 : determining the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types.

[0158] In the embodiment of the present disclosure, the failure probabilities of multiple sub-fault types can be determined according to the index values ​​of the multiple sub-fault types. For example, for each sub-fault type, there is a positive relationship between the index value of the sub-fault type and the failure probability, that is, the larger the index value of the sub-fault type, the greater the failure probability of the sub-fault type, and conversely, the smaller the index value of the sub-fault type, the smaller the failure probability of the sub-fault type.

[0159] As a possible implementation manner, the second coefficient may be determined according to the sum of index values ​​of multiple sub-fault types.

[0160] For example, the indicator values ​​marking the m seed fault types are CRI1, CRI2, ..., CRI m , that is, the index value of the i-th seed fault type is CRI i , the sum of the index values ​​of the above m seed fault types is Then the second coefficient can be

[0161] After the first coefficient is determined, the failure probability of each sub-fault type may be determined according to the ratio of the index value of each sub-fault type to the second coefficient.

[0162] For example, still using the above example, the index value of the i-th fault type is CRI i , the second coefficient is The failure probability of the i-th seed failure type can be:

[0163] Step 207 : determining a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types.

[0164] The execution process of step 207 may refer to the execution process of any embodiment of the present disclosure, and will not be described in detail here.

[0165] The fault identification method of the gas turbine of the embodiment of the present disclosure generates a fault vector corresponding to the i-th sub-fault type according to the i-th row element in the target fault identification matrix; determines the similarity between the fault vector and the feature vector of the i-th sub-fault type; and determines the index value of the i-th sub-fault type according to the similarity. Thus, the index value of the corresponding sub-fault type is effectively determined according to the similarity between the feature vector related to the target fault type obtained by monitoring the gas turbine and the fault vector corresponding to each sub-fault type in the target fault identification matrix.

[0166] Based on any of the above embodiments, in order to clearly illustrate how to generate a feature vector related to a specified target fault type based on monitoring values ​​of multiple dimensions, the present disclosure also proposes a gas turbine fault identification method.

[0167] Figure 3 A flow chart of a gas turbine fault identification method provided in Embodiment 3 of the present disclosure.

[0168] like Figure 3 As shown, the gas turbine fault identification method may include the following steps:

[0169] Step 301: monitor the gas turbine to obtain monitoring values ​​in multiple dimensions.

[0170] The execution process of step 301 can refer to the execution process of any embodiment of the present disclosure, and will not be described in detail here.

[0171] It can be understood that after obtaining the monitoring values ​​of multiple dimensions, the monitoring values ​​of multiple dimensions can be preprocessed, wherein the preprocessing can include at least one of missing value processing, outlier processing, duplicate value processing and noise interference processing to improve the accuracy and completeness of the monitoring values; and / or, the preprocessing can include normalization processing or standardization processing, so as to facilitate subsequent data processing.

[0172] Step 302: Determine a target value related to a target fault type from monitoring values ​​of multiple dimensions.

[0173] In the embodiment of the present disclosure, a target value related to a target fault type may be determined from monitoring values ​​of multiple dimensions.

[0174] As an example, the monitoring values ​​of multiple dimensions may include compressor inlet temperature, compressor outlet temperature, compressor inlet pressure, compressor outlet pressure, compressor speed, compressor inlet air flow, compressor outlet air flow, oil temperature at the lubricating oil system inlet, oil temperature at the lubricating oil system outlet, bearing vibration frequency, bearing vibration speed, bearing vibration acceleration, turbine inlet pressure, turbine outlet pressure, turbine inlet air flow, turbine outlet air flow and other monitoring values. When the target fault type is an air path fault, the target value related to the air path fault can be determined from the monitoring values ​​of the above multiple dimensions. For example, the target value includes the monitoring values ​​corresponding to various parameters such as compressor inlet temperature, compressor outlet temperature, compressor inlet pressure, compressor outlet pressure, compressor inlet air flow, compressor outlet air flow, turbine inlet pressure, turbine outlet pressure, turbine inlet air flow, turbine outlet air flow and other parameters.

[0175] It should be noted that the above example of determining the target value related to the gas circuit fault from the monitoring values ​​of multiple dimensions is only exemplary. In actual applications, the target value related to the gas circuit fault can be determined from the monitoring of multiple dimensions according to actual needs and application scenarios. For example, the target value can also include the monitoring values ​​corresponding to parameters such as the power turbine inlet pressure and the power turbine outlet pressure.

[0176] It should also be noted that in the present disclosure, not only can the monitoring values ​​of multiple dimensions be preprocessed after obtaining the monitoring values ​​of multiple dimensions, but also after obtaining the monitoring values ​​of multiple dimensions and determining the target value related to the target fault type from each monitoring value, each target value can be preprocessed. The preprocessing method is the same as described above and will not be elaborated here.

[0177] Step 303: Map each target value to obtain a measurement value corresponding to each performance parameter related to the target fault type.

[0178] In the embodiment of the present disclosure, the target fault type may have various related performance parameters. For example, taking the target fault type as a gas line fault as an example, the performance parameters related to the gas line fault may include a flow parameter and an efficiency parameter.

[0179] It should be noted that, when the gas turbine does not include a power turbine, the performance parameters related to the gas path failure may include compressor flow, compressor efficiency, turbine flow, and turbine efficiency; and when the gas turbine includes a power turbine, the performance parameters related to the gas path failure may include compressor flow, compressor efficiency, turbine flow, turbine efficiency, power turbine flow, and power turbine efficiency.

[0180] In the embodiments of the present disclosure, each target value may be mapped according to a mapping formula in the related art, thereby obtaining a measurement value corresponding to each performance parameter related to the target fault type.

[0181] For example, there is a mapping relationship between the compressor inlet air flow, the compressor outlet air flow, the compressor inlet pressure and the compressor outlet pressure, and the compressor efficiency, a performance parameter related to the air path failure, that is, the measured value corresponding to the compressor efficiency can be determined according to the target value corresponding to the compressor inlet air flow, the target value corresponding to the compressor outlet air flow, the target value corresponding to the compressor inlet pressure and the target value corresponding to the compressor outlet pressure.

[0182] Step 304, obtaining reference values ​​corresponding to various performance parameters, wherein the gas turbine operates under the reference values ​​without any fault corresponding to the target fault type.

[0183] In the embodiment of the present disclosure, each performance parameter may have a corresponding reference value, and the gas turbine operates at the reference value without causing a fault corresponding to the target fault type.

[0184] It should be noted that the reference value can be determined based on the monitoring values ​​of multiple dimensions of the gas turbine monitored when the gas turbine does not have a fault corresponding to the target fault type, that is, when the gas turbine operates normally, that is, based on the monitoring values ​​of multiple dimensions obtained when the gas turbine operates normally, the performance values ​​corresponding to each performance parameter related to the target fault type are determined, and each performance value is mapped to obtain the reference value corresponding to each performance parameter.

[0185] Step 305: Generate a characteristic value corresponding to each performance parameter according to the difference between the measured value corresponding to each performance parameter and the reference value.

[0186] In the embodiment of the present disclosure, the characteristic value corresponding to each performance parameter can be generated according to the difference between the measured value and the reference value corresponding to each performance parameter, wherein the difference can be the difference between the measured value and the reference value corresponding to each performance parameter.

[0187] For example, based on the difference between the measured value and the reference value corresponding to each performance parameter in the gas turbine, the generated characteristic value corresponding to the compressor flow is x1-y1, the characteristic value corresponding to the compressor efficiency is x2-y2, the characteristic value corresponding to the turbine flow is x3-y3, and the characteristic value corresponding to the turbine efficiency is x4-y4.

[0188] Among them, x1 represents the measured value corresponding to the compressor flow, and y1 represents the reference value corresponding to the compressor flow; x2 represents the measured value corresponding to the compressor efficiency, and y2 represents the reference value corresponding to the compressor efficiency; x3 represents the measured value corresponding to the turbine flow, and y3 represents the reference value corresponding to the turbine flow; x4 represents the measured value corresponding to the turbine efficiency, and y4 represents the reference value corresponding to the turbine efficiency.

[0189] Step 306: Generate a feature vector according to the feature value of each performance parameter.

[0190] In the disclosed embodiment, a feature vector may be generated according to the feature value of each performance parameter.

[0191] Still taking the above example, according to the performance parameters related to the gas path failure in the gas turbine, namely, the characteristic value corresponding to the compressor flow (x1-y1), the characteristic value corresponding to the compressor efficiency (x2-y2), the characteristic value corresponding to the turbine flow (x3-y3), and the characteristic value corresponding to the turbine efficiency (x4-y4), the generated characteristic vector is, for example, (x1-y1, x2-y2, x3-y3, x4-y4)′.

[0192] It should be noted that the above example of sorting the performance parameters in the feature vector is only illustrative. In practical applications, the performance parameters can be sorted as needed. For example, still taking the above example as an example, the generated feature vector can be (x4-y4,x3-y3,x2-y2,x1-y1)′.

[0193] Step 307 , querying a target fault identification matrix that matches the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector.

[0194] Step 308: determining the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types.

[0195] Step 309 : determining a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types.

[0196] The execution process of steps 307 to 309 may refer to the execution process of any embodiment of the present disclosure, and will not be described in detail here.

[0197] The fault identification method of the gas turbine of the embodiment of the present disclosure determines the target value related to the target fault type from the monitoring values ​​of multiple dimensions; maps each target value to obtain the measurement value corresponding to each performance parameter related to the target fault type; obtains the reference value corresponding to each performance parameter, wherein the gas turbine does not have the fault corresponding to the target fault type when operating under the reference value; generates the characteristic value corresponding to each performance parameter according to the difference between the measurement value and the reference value corresponding to each performance parameter; generates the characteristic vector according to the characteristic value of each performance parameter. Thus, the characteristic vector related to the target fault type can be effectively determined according to the monitoring values ​​of multiple dimensions.

[0198] As an example, the fault identification process of the gas turbine proposed in the present disclosure can be as follows: Figure 4 As shown, it may include the following steps:

[0199] 1. Obtain various monitoring values ​​of the gas turbine

[0200] By monitoring the gas turbine, it is possible to obtain monitoring values ​​of the gas turbine in multiple dimensions, wherein the monitoring values ​​of the gas turbine may include various types of data such as vibration, performance, and lubricating oil.

[0201] 2. Classification and preprocessing of each monitoring value

[0202] After obtaining each monitoring value, each monitoring value can be classified, and the target value related to the specified target fault identification type (such as gas circuit fault and / or structural strength fault) can be screened out from each monitoring value. After determining the target value related to the target fault identification type, each target value can be cleaned, that is, the abnormal values ​​in each target value can be cleaned. The abnormal value cleaning can include missing value processing, outlier processing, duplicate value processing and noise interference processing, etc., to improve the integrity and accuracy of each target value obtained, and each target value can be pre-processed, such as normalization processing or standardization processing, to facilitate subsequent data processing.

[0203] 3. Constructing comprehensive ranking indicators

[0204] Combining the advantages of the distance measurement algorithm and the similarity measurement algorithm, that is, combining the advantages of the distance measurement algorithm in identifying the sample vector modulus length and the similarity measurement algorithm in identifying the sample vector direction, a comprehensive ranking index (Comprehensive Ranking Index, CRI) that takes into account both the sample modulus length and the sample vector is constructed. The index can be calculated according to the following formula:

[0205]

[0206] Among them, SM can be a distance obtained by using any algorithm in the distance measurement algorithm, and DM can be a similarity obtained by using any algorithm in the similarity measurement algorithm.

[0207] That is, when constructing a comprehensive ranking index, it is necessary to determine one of the distance measurement algorithms to be used and one of the similarity measurement algorithms to be used.

[0208] 4. Determine the comprehensive ranking index of various sub-fault types

[0209] 4.1 Map each target value to obtain the measured value corresponding to each performance parameter related to the target fault type, and obtain the reference value of each performance parameter. According to the difference between the measured value and the reference value corresponding to each performance parameter, generate the feature vector related to the target fault type.

[0210] For example, the measured values ​​corresponding to the performance parameters related to the gas path fault in the gas turbine are: compressor flow x1, compressor efficiency x2, turbine flow x3, turbine efficiency x4, and the reference values ​​corresponding to the performance parameters related to the gas path fault in the gas turbine are: compressor flow y1, compressor efficiency y2, turbine flow y3, turbine efficiency y4, and the differences between the measured values ​​and the reference values ​​corresponding to the performance parameters in the gas turbine are: x1-y1, x2-y2, x3-y3, x4-y4, so that the generated feature vector related to the target fault type can be (x1-y1, x2-y2, x3-y3, x4-y4)′, which reflects the changes in the performance parameters related to the gas path fault in the gas turbine.

[0211] 4.2 Obtaining the target fault identification matrix

[0212] For example, the target fault type is gas path fault, and the gas turbine does not include the power turbine. The sub-fault types under the gas path fault may include compressor blade fouling (marked as fault 1), compressor blade wear (marked as fault 2), compressor blade mechanical damage (marked as fault 3), compressor blade tip clearance increase (marked as fault 4), turbine blade fouling (marked as fault 5), turbine blade wear (marked as fault 6), turbine blade mechanical damage (marked as fault 7), turbine blade thermal corrosion (marked as fault 8), and the performance parameter set is (compressor flow, compressor efficiency, turbine flow, turbine efficiency). According to manual experience, when the gas turbine has fault 1, the fault vectors corresponding to faults 1 to 8 can be generated according to Table 1. Therefore, according to the fault vectors of each sub-fault, the corresponding fault identification matrix can be generated as follows:

[0213]

[0214] 4.3 Based on the target fault identification matrix and the above eigenvectors, generate the index values ​​of multiple sub-fault types under the target fault type

[0215] The fault vector corresponding to the i-th seed fault type can be generated according to the i-th row element in the target fault identification matrix; wherein i is a positive integer not greater than m, and m is the number of sub-fault types under the target fault; the similarity between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by the similarity measurement algorithm determined in step 3, and the distance between the fault vector of the i-th seed fault type and the above-mentioned feature vector is determined by the distance measurement algorithm determined in step 3; according to the similarity and the distance, the index value CRI of the i-th seed fault type is determined according to formula (16): i .

[0216] According to the above steps, the index values ​​corresponding to all sub-fault types under the target fault type can be determined, so that the comprehensive ranking index vector of the target fault can be (CRI1, CRI2,…, CRI m )′.

[0217] 5. Determine the failure probability of each sub-fault type of the gas turbine

[0218] Based on the comprehensive ranking index vector of the target fault, the failure probability of each sub-fault type under the target fault can be determined according to the following formula. For example, the failure probability of the i-th sub-fault is:

[0219]

[0220] 6. Determine the target sub-fault type by automatically sorting the fault probability

[0221] Combined with the automatic sorting algorithm, that is, arranging the sub-fault types in descending order according to the fault probability of each sub-fault type, for example, the sub-fault type with a fault probability greater than 50% can be selected as the target sub-fault type, and the target sub-fault type is the fault type to which the gas turbine belongs.

[0222] It should be noted that after determining the target fault type to which the gas turbine belongs, when the target sub-fault type is different from the sub-fault type that occurs in the gas turbine in the actual scenario, the indicator in step 3 can be replaced, that is, the distance measurement algorithm is replaced (for example, the ED algorithm is replaced with the MAD algorithm), and / or the similarity algorithm is replaced (for example, the PCC algorithm is replaced with the CS algorithm), until the optimal distance measurement algorithm and / or the optimal similarity algorithm are found, so that the fault probability of each sub-fault type can be recalculated according to the optimal distance measurement algorithm and the optimal similarity algorithm, so that the target sub-fault type selected according to the fault probability of each sub-fault type is the same as the sub-fault type that occurs in the gas turbine in the actual scenario. Therefore, when the fault identification method disclosed in the present invention is used to identify the next fault of the gas turbine, the fault type to which the gas turbine belongs can be identified based on the above-mentioned optimal distance measurement algorithm and / or the optimal similarity algorithm.

[0223] As an example, the inventors use formula (16) to calculate the pattern similarity between eight sub-fault types under typical gas path faults. The calculation results can be shown in Table 4:

[0224] Table 4 Calculation results based on comprehensive ranking indicators

[0225] Sub-fault category Fault 1 Fault 2 Fault 3 Fault 4 Fault 5 Fault 6 Fault 7 Fault 8 Fault 1 1.000 0.028 0.028 0.598 -0.106 0.051 -0.079 0.088 Fault 2 0.028 1.000 1.000 -0.061 -0.082 0.039 -0.061 0.066 Fault 3 0.028 1.000 1.000 -0.061 -0.082 0.039 -0.061 0.066 Fault 4 0.598 -0.061 -0.061 1.000 -0.082 0.039 -0.061 0.066 Fault 5 -0.106 -0.082 -0.082 -0.082 1.000 -0.234 0.044 -0.272 Fault 6 0.051 0.039 0.039 0.039 -0.234 1.000 0.173 0.591 Fault 7 -0.079 -0.061 -0.061 -0.061 0.044 0.173 1.000 0.066 Fault 8 0.088 0.066 0.066 0.066 -0.272 0.591 0.066 1.000

[0226] Among them, the CRI value between fault i (i is a positive integer not greater than 8) and fault j (j is a positive integer not greater than 8) in Table 4 refers to the pattern similarity between the fault criterion corresponding to fault i and the fault criterion corresponding to fault j calculated using formula (16).

[0227] It can be seen from Table 4 that the pattern similarity (CRI value) between the fault criteria of fault 1 and the fault criteria of fault 4, and the pattern similarity (CRI value) between the fault criteria of fault 6 and the fault criteria of fault 8 are reduced to below 0.6, which can effectively distinguish fault 4 from fault 6, and effectively distinguish fault 6 from fault 8.

[0228] It should be noted that the above only takes the target fault type as a gas circuit fault as an example, but the present disclosure is not limited thereto, and the target fault type may be a gas circuit fault and / or a structural strength fault.

[0229] In summary, by real-time monitoring of the characteristics of performance parameters related to the specified target fault type in the gas turbine, learning the changes in performance parameters, combining the advantages of distance measurement algorithm and similarity measurement algorithm, constructing a comprehensive sorting index, and determining the fault probability based on the comprehensive sorting index vector, the fault type of the gas turbine can be identified based on the automatic sorting of the fault probability.

[0230] The gas turbine fault identification method based on the embodiment of the present disclosure can at least embody its advantages in the following aspects:

[0231] 1) When constructing the comprehensive ranking index, the module length and direction characteristics of the sample vector are taken into account, combining the advantages of the distance measurement algorithm and the similarity measurement algorithm;

[0232] (2) The fault identification matrix obtained through prior knowledge can improve the identification efficiency and ensure the timeliness of fault identification;

[0233] (3) An automatic identification method for gas turbine faults is proposed, which can provide technical support for the expert diagnosis system;

[0234] (4) This method is not limited to the automatic identification of gas turbine faults, but also has migration value.

[0235] With the above Figures 1 to 3 Corresponding to the gas turbine fault identification method provided in the embodiment, the present disclosure also provides a gas turbine fault identification device. Figures 1 to 3 The gas turbine fault identification method provided in the embodiment corresponds to the embodiment, so the implementation method of the gas turbine fault identification method is also applicable to the gas turbine fault identification device provided in the embodiment of the present disclosure, and will not be described in detail in the embodiment of the present disclosure.

[0236] Figure 5 This is a schematic diagram of the structure of a gas turbine fault identification device provided in Embodiment 4 of the present disclosure.

[0237] like Figure 5 As shown, the gas turbine fault identification device 500 may include: a monitoring module 501 , a generating module 502 , a processing module 503 , a first determining module 504 and a second determining module 505 .

[0238] The monitoring module 501 is used to monitor the gas turbine to obtain monitoring values ​​in multiple dimensions.

[0239] The generating module 502 is used to generate a feature vector related to a specified target fault type according to monitoring values ​​of multiple dimensions.

[0240] The processing module 503 is used to query the target fault identification matrix that matches the target fault type, and generate index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector.

[0241] The first determination module 504 is configured to determine the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types.

[0242] The second determination module 505 is used to determine a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types.

[0243] In a possible implementation of the embodiment of the present disclosure, the number of elements included in the feature vector is n, the number of sub-fault types under the target fault type is m, the target fault identification matrix is ​​an m*n matrix, and m and n are both positive integers; the processing module 503 is specifically used to: generate a fault vector corresponding to the i-th seed fault type according to the i-th row elements in the target fault identification matrix; wherein i is a positive integer not greater than m; determine the similarity between the fault vector and the feature vector of the i-th seed fault type; and determine the index value of the i-th seed fault type according to the similarity.

[0244] In a possible implementation of the embodiment of the present disclosure, the processing module 503 is specifically used to: determine the distance between the fault vector and the characteristic vector of the i-th seed fault type; determine the first coefficient based on the sum of the similarity and the distance; determine the index value of the i-th seed fault type based on the ratio of the similarity to the first coefficient.

[0245] In a possible implementation of the embodiment of the present disclosure, the first determination module 504 is specifically used to: determine the second coefficient according to the sum of the index values ​​of multiple sub-fault types; and determine the failure probability of each sub-fault type according to the ratio of the index value of each sub-fault type to the second coefficient.

[0246] In a possible implementation of the embodiment of the present disclosure, the second determination module 505 is specifically used to: determine the sub-fault type with a failure probability greater than a set probability threshold as the target sub-fault type to which the gas turbine belongs; or determine the sub-fault type corresponding to the maximum failure probability as the target sub-fault type to which the gas turbine belongs.

[0247] In a possible implementation of the embodiment of the present disclosure, the generation module 502 is specifically used to: determine a target value related to a target fault type from monitoring values ​​of multiple dimensions; map each target value to obtain a measurement value corresponding to each performance parameter related to the target fault type; obtain a reference value corresponding to each performance parameter, wherein the gas turbine operates at the reference value without a fault corresponding to the target fault type; generate a characteristic value corresponding to each performance parameter based on the difference between the measurement value and the reference value corresponding to each performance parameter; and generate a characteristic vector based on the characteristic value of each performance parameter.

[0248] In a possible implementation of the embodiment of the present disclosure, the gas turbine fault identification device 500 may further include:

[0249] A preprocessing module is used to preprocess the monitoring values ​​of multiple dimensions, wherein the preprocessing includes at least one of missing value processing, outlier processing, duplicate value processing and noise interference processing, and / or the preprocessing includes normalization processing or standardization processing.

[0250] The fault identification device of the gas turbine of the disclosed embodiment monitors the gas turbine to obtain monitoring values ​​of multiple dimensions; generates a characteristic vector related to the specified target fault type according to the monitoring values ​​of multiple dimensions; queries the target fault identification matrix matching the target fault type, and generates index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector; determines the fault probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types; and determines the target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types. Thus, the probability of each sub-fault type under the target fault type occurring in the gas turbine can be effectively determined according to the monitoring values ​​of multiple dimensions obtained by monitoring the gas turbine, and then the fault type to which the gas turbine belongs can be effectively identified based on the probability of each sub-fault type, thereby improving the accuracy and reliability of the fault identification result.

[0251] In order to implement the above embodiments, the present disclosure further proposes an electronic device, wherein the electronic device can be the server or detection device in the aforementioned embodiments; it includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the program, it implements the gas turbine fault identification method proposed in any of the aforementioned embodiments of the present disclosure.

[0252] In order to implement the above embodiments, the present disclosure also proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the gas turbine fault identification method proposed in any of the above embodiments of the present disclosure is implemented.

[0253] In order to implement the above embodiments, the present disclosure further proposes a computer program product. When the instructions in the computer program product are executed by a processor, the gas turbine fault identification method proposed in any of the above embodiments of the present disclosure is executed.

[0254] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0255] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the description of the present disclosure, "plurality" means at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0256] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present disclosure belong.

[0257] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0258] It should be understood that the various parts of the present disclosure can be implemented in hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0259] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0260] In addition, each functional unit in each embodiment of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0261] The storage medium mentioned above may be a read-only memory, a disk or an optical disk, etc. Although the embodiments of the present disclosure have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations of the present disclosure. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present disclosure.

Claims

1. A method for identifying a fault of a gas turbine, characterized in that: The method comprises: Monitoring the gas turbine to obtain monitoring values ​​in multiple dimensions; Generating a feature vector related to a specified target fault type according to the monitoring values ​​of the multiple dimensions; Querying a target fault identification matrix matching the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector; Determining the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types; Determining a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types; The number of elements included in the feature vector is n, the number of sub-fault types under the target fault type is m, and the target fault identification matrix is ​​an m*n matrix, where m and n are both positive integers; The step of generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector includes: Generate a fault vector corresponding to the i-th seed fault type according to the i-th row elements in the target fault identification matrix; wherein i is a positive integer not greater than m; Determining the similarity between the fault vector of the i-th seed fault type and the feature vector; Determining an index value of the i-th seed fault type according to the similarity; Determining the index value of the i-th seed fault type according to the similarity includes: Determining a distance between a fault vector of the i-th fault type and the feature vector; Determine a first coefficient according to the sum of the similarity and the distance; Determining an index value of the i-th seed fault type according to a ratio of the similarity to the first coefficient; The determining the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types includes: Determining a second coefficient according to the sum of the index values ​​of the plurality of sub-fault types; The failure probability of each of the sub-fault types is determined according to the ratio of the index value of each of the sub-fault types and the second coefficient.

2. The method according to claim 1, characterized in that The step of determining the target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types comprises: Determine the sub-fault type whose fault probability is greater than a set probability threshold as the target sub-fault type to which the gas turbine belongs; or, The sub-fault type corresponding to the maximum fault probability is determined as the target sub-fault type to which the gas turbine belongs.

3. The method according to any one of claims 1 to 2, characterized in that: The step of generating a feature vector related to a specified target fault type according to the monitoring values ​​of the multiple dimensions includes: Determining a target value associated with the target fault type from the monitoring values ​​of the multiple dimensions; Mapping each of the target values ​​to obtain a measurement value corresponding to each performance parameter associated with the target fault type; Acquiring reference values ​​corresponding to the performance parameters, wherein the gas turbine operates under the reference values ​​without causing a fault corresponding to the target fault type; generating a characteristic value corresponding to each of the performance parameters according to a difference between a measured value corresponding to each of the performance parameters and a reference value; The feature vector is generated according to the feature value of each performance parameter.

4. The method according to claim 3, characterized in that: Before determining the target value related to the target fault type from the monitoring values ​​of the multiple dimensions, the method further includes: The monitoring values ​​of the multiple dimensions are preprocessed, wherein the preprocessing includes at least one of missing value processing, outlier processing, duplicate value processing and noise interference processing, and / or the preprocessing includes normalization processing or standardization processing.

5. A gas turbine fault identification device, characterized in that: The device comprises: A monitoring module, used for monitoring the gas turbine to obtain monitoring values ​​in multiple dimensions; A generating module, used for generating a feature vector related to a specified target fault type according to the monitoring values ​​of the multiple dimensions; A processing module, used for querying a target fault identification matrix matching the target fault type, and generating index values ​​of multiple sub-fault types under the target fault type according to the target fault identification matrix and the characteristic vector; A first determination module, configured to determine the failure probabilities of the multiple sub-fault types according to the index values ​​of the multiple sub-fault types; A second determination module is used to determine a target sub-fault type to which the gas turbine belongs from the multiple sub-fault types according to the fault probabilities of the multiple sub-fault types; The number of elements included in the feature vector is n, the number of sub-fault types under the target fault type is m, and the target fault identification matrix is ​​an m*n matrix, where m and n are both positive integers; The processing module is specifically used for: Generate a fault vector corresponding to the i-th seed fault type according to the i-th row elements in the target fault identification matrix; wherein i is a positive integer not greater than m; Determining the similarity between the fault vector of the i-th seed fault type and the feature vector; Determining an index value of the i-th seed fault type according to the similarity; The processing module is specifically used for: Determining a distance between a fault vector of the i-th fault type and the feature vector; Determine a first coefficient according to the sum of the similarity and the distance; Determining an index value of the i-th seed fault type according to a ratio of the similarity to the first coefficient; The first determining module is specifically configured to: Determining a second coefficient according to the sum of the index values ​​of the plurality of sub-fault types; The failure probability of each of the sub-fault types is determined according to the ratio of the index value of each of the sub-fault types and the second coefficient.

6. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 4 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images