Fault diagnosis method, device and equipment for turbine cooler

Abstract

The invention provides a fault diagnosis method, device and equipment for a turbine cooler. The method provided by the invention comprises the following steps: acquiring real-time working condition data and a real-time state signal of the turbine cooler; respectively carrying out feature extraction on the real-time working condition data and the real-time state signal to obtain a real-time working condition feature and a real-time state feature; performing reduction processing on the real-time working condition feature and the real-time state feature to obtain a first simplified feature and a second simplified feature; performing dimension reduction processing on the first simplified feature and the second simplified feature to obtain a first dimension reduction feature and a second dimension reduction feature; and combining the first dimension reduction feature and the second dimension reduction feature to obtain a combined feature, and inputting the combined feature into a pre-trained fault diagnosis model, so that the fault diagnosis model identifies the fault of the turbine cooler based on the combined feature. The turbine cooler fault diagnosis method, device and equipment provided by the invention are used for accurately performing fault diagnosis of the turbine cooler.

Description

Technical Field

[0001] The present application relates to the technical field of fault diagnosis, and in particular to a fault diagnosis method, device and equipment for a turbine cooler. Background Art

[0002] Turbine coolers have become key equipment in many industries due to their efficient heat dissipation and energy efficiency. In the aerospace field, advanced turbine cooling technology has effectively improved the performance and high temperature resistance of turbine engines, making them adaptable to a wider range of working environments. Turbine coolers operate in high energy consumption and high temperature environments, and their failures may cause significant economic losses and safety hazards. Therefore, how to accurately diagnose turbine cooler faults is a hot topic in current research. Summary of the invention

[0003] In view of this, the present application provides a fault diagnosis method, device and equipment for a turbine cooler, so as to accurately perform fault diagnosis of the turbine cooler.

[0004] Specifically, the present application is implemented through the following technical solutions: A first aspect of the present application provides a fault diagnosis method for a turbine cooler, the method comprising: Obtain real-time operating data and real-time status signals of the turbine cooler; Performing feature extraction on the real-time operating condition data and the real-time status signal respectively to obtain real-time operating condition features corresponding to the real-time operating condition data and real-time status features corresponding to the real-time status signal; Simplifying the real-time working condition feature and the real-time state feature respectively to obtain a first simplified feature corresponding to the real-time working condition feature and a second simplified feature corresponding to the real-time state feature; Performing dimensionality reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimensionality reduction feature corresponding to the first simplified feature and a second dimensionality reduction feature corresponding to the second simplified feature; The first dimensionality reduction feature and the second dimensionality reduction feature are merged to obtain a merged feature, and the merged feature is input into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature; wherein the dimensionality difference between the first dimensionality reduction feature and the second dimensionality reduction feature is less than a preset range, and the fault diagnosis model can pay attention to the first dimensionality reduction feature and the second dimensionality reduction feature at the same time, and perform fault diagnosis by combining the first dimensionality reduction feature and the second dimensionality reduction feature.

[0005] A second aspect of the present application provides a fault diagnosis device for a turbine cooler, the device comprising an acquisition module, an extraction module, a simplification module, a dimensionality reduction module and a processing module; wherein: The acquisition module is used to acquire real-time operating condition data and real-time status signals of the turbine cooler; The extraction module is used to extract features from the real-time operating condition data and the real-time status signal respectively, to obtain the real-time operating condition features corresponding to the real-time operating condition data and the real-time status features corresponding to the real-time status signal; The simplification module is used to simplify the real-time working condition feature and the real-time state feature respectively to obtain a first simplified feature corresponding to the real-time working condition feature and a second simplified feature corresponding to the real-time state feature; The dimension reduction module is used to perform dimension reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimension reduction feature corresponding to the first simplified feature and a second dimension reduction feature corresponding to the second simplified feature; The processing module is used to merge the first dimensionality reduction feature and the second dimensionality reduction feature to obtain a merged feature, and input the merged feature into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature; wherein the dimensional difference between the first dimensionality reduction feature and the second dimensionality reduction feature is less than a preset range, and the fault diagnosis model can pay attention to the first dimensionality reduction feature and the second dimensionality reduction feature at the same time, and perform fault diagnosis by combining the first dimensionality reduction feature and the second dimensionality reduction feature.

[0006] A third aspect of the present application provides a fault diagnosis device for a turbine cooler, comprising 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 steps of any one of the methods provided in the first aspect of the present application are implemented.

[0007] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods provided in the first aspect of the present application.

[0008] The fault diagnosis method, device and equipment of the turbine cooler provided in the present application obtain the real-time operating condition data and real-time status signal of the turbine cooler, and perform feature extraction on the real-time operating condition data and the real-time status signal to obtain the real-time operating condition characteristics corresponding to the real-time operating condition data, and the real-time status characteristics corresponding to the real-time status signal, and then simplify the real-time operating condition characteristics and the real-time status characteristics respectively to obtain the first simplified characteristics corresponding to the real-time operating condition characteristics, and the second simplified characteristics corresponding to the real-time status characteristics, and perform dimensionality reduction processing on the first simplified characteristics and the second simplified characteristics respectively to obtain the first reduced dimensionality characteristics corresponding to the first simplified characteristics, and the second reduced dimensionality characteristics corresponding to the second simplified characteristics, and then merge the first reduced dimensionality characteristics and the second reduced dimensionality characteristics to obtain the merged characteristics, and input the merged characteristics into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged characteristics. In this way, on the first hand, by performing feature extraction, simplification processing and dimensionality reduction processing on the real-time working condition data and the real-time state signal respectively, the dimension difference between the dimensionality reduction features corresponding to the working condition data and the dimensionality reduction features corresponding to the state signal is smaller than the preset range. In this way, not only can the working condition data and the state information be combined for fault diagnosis at the same time, but also the negotiation between the two in the feature dimension can be ensured, ensuring that the fault diagnosis model can simultaneously and balancedly consider the working condition characteristics and the state characteristics in the process of fault diagnosis, and combine the working condition characteristics and the state characteristics to perform fault diagnosis, so as to accurately capture the fault characteristics related to the working condition changes, and make a comprehensive judgment in combination with the state characteristics. In this way, the accuracy and reliability of fault diagnosis can be improved, and the adaptability of the fault diagnosis model to different working conditions can be enhanced; on the second hand, through feature simplification and dimensionality reduction, the dimension of the original feature data is reduced, the redundant information and the computational complexity are reduced, so that the dimension difference between the final first dimensionality reduction feature and the second dimensionality reduction feature is smaller than the preset range, which helps the fault diagnosis model to effectively focus on key features and avoid the model overfitting problem caused by high-dimensional data. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flow chart of a first embodiment of a fault diagnosis method for a turbine cooler provided in the present application; Figure 2 A flow chart of a second embodiment of a fault diagnosis method for a turbine cooler provided in the present application; Figure 3 A schematic diagram of the weights of various sub-features in a state feature shown in an exemplary embodiment of the present application; Figure 4 A schematic diagram showing the information ratio of each component included in a simplified feature of an exemplary embodiment provided by the present application; Figure 5 A schematic diagram of a confusion matrix of a fault diagnosis model provided in this application; Figure 6 A hardware structure diagram of a fault diagnosis device of a turbine cooler where the fault diagnosis device of the turbine cooler of the present application is located; Figure 7 This is a schematic structural diagram of Embodiment 1 of the fault diagnosis device for a turbine cooler provided in the present application. DETAILED DESCRIPTION

[0010] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0011] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to and includes any or all possible combinations of one or more associated listed items.

[0012] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0013] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0014] Figure 1 This is a flow chart of the first embodiment of the fault diagnosis method for a turbine cooler provided in this application. Figure 1 , the method provided in this embodiment may include: S101. Acquire real-time operating data and real-time status signals of a turbine cooler.

[0015] Specifically, a turbine cooler is a device used to cool a turbine or other mechanical equipment, which absorbs and removes the heat generated during the operation of the turbine through a circulating cooling medium (such as water, air or coolant).

[0016] Furthermore, the real-time operating condition data is data collected in real time that reflects the working conditions of the turbine cooler. It should be noted that the specific data type included in the real-time operating condition data is determined according to actual needs, and in this embodiment, it is not set. In specific implementation, the temperature data, pressure data, flow data, speed data and other data of the turbine cooler collected in real time can be determined as the real-time operating condition data.

[0017] Furthermore, the real-time status signal refers to a signal collected in real time by a monitoring device such as a sensor, which reflects the current operating status of the turbine cooler. It should be noted that the specific signal type included in the real-time status signal is determined according to actual needs, and is not set in this embodiment. In specific implementation, the vibration signal, noise signal, temperature fluctuation signal and other signals of the turbine cooler collected in real time can be determined as the real-time status signal.

[0018] It should be noted that, referring to the previous description, it can be understood that the operating condition data mainly describes the operating parameters of the equipment under specific operating conditions, which usually include environmental or operating variables that directly affect the operation of the equipment; while the status signal is used to describe the health status or performance status of the equipment, reflecting whether there is any abnormality or degradation in the equipment, which is usually related to the structural health and reliability of the equipment, such as vibration, noise, etc.

[0019] In specific implementation, various sensors such as temperature sensors, pressure sensors, flow sensors, speed sensors, and vibration sensors can be installed at key locations of the turbine cooler to collect real-time operating data and real-time status signals of the turbine cooler. In this step, the corresponding data is directly obtained from these sensors.

[0020] S102 , performing feature extraction on the real-time operating condition data and the real-time status signal respectively to obtain real-time operating condition features corresponding to the real-time operating condition data and real-time status features corresponding to the real-time status signal.

[0021] Specifically, feature extraction is the process of extracting useful information from raw data. It is understandable that the feature representation obtained through feature extraction can be used for subsequent data analysis or machine learning tasks.

[0022] It should be noted that the real-time operating condition features and real-time status features can be obtained based on traditional feature extraction methods, or based on neural network feature extraction methods. This is not limited in this embodiment.

[0023] Furthermore, the real-time operating condition feature is a feature representation obtained from the real-time operating condition data through feature extraction, and the real-time state feature is a feature representation obtained from the real-time state signal through feature extraction.

[0024] S103 , simplifying the real-time operating condition feature and the real-time state feature respectively to obtain a first simplified feature corresponding to the real-time operating condition feature and a second simplified feature corresponding to the real-time state feature.

[0025] Specifically, the reduction process is used to reduce the amount of data of the real-time working condition feature and the real-time state feature, while retaining the information of the original data as much as possible. It can be understood that the reduction process can improve the efficiency and accuracy of subsequent data processing by removing redundant or unimportant features.

[0026] It should be noted that the real-time operating condition characteristics and real-time state characteristics can be simplified based on the traditional simplification processing method, or based on the neural network simplification processing method, which is not limited in this embodiment. In specific implementation, the relieff algorithm can be used to complete the simplification process to obtain the first simplified feature and the second simplified feature.

[0027] Furthermore, the real-time operating condition feature is simplified to obtain a first simplified feature; and the real-time state feature is simplified to obtain a second simplified feature.

[0028] S104: Perform dimensionality reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimensionality reduction feature corresponding to the first simplified feature and a second dimensionality reduction feature corresponding to the second simplified feature.

[0029] Specifically, dimensionality reduction is used to map high-dimensional data to low-dimensional space through mathematical transformation, while retaining the structure and information of the data as much as possible. It can be understood that dimensionality reduction helps to reduce computational complexity, improve data processing efficiency, and possibly avoid overfitting problems.

[0030] It should be noted that the dimensionality reduction can be performed based on a traditional dimensionality reduction method or a neural network dimensionality reduction method, which is not limited in this embodiment. In specific implementation, the principal component analysis method can be used to complete the dimensionality reduction process to obtain the first dimensionality reduction feature and the second dimensionality reduction feature.

[0031] Further, the first simplified feature is subjected to dimensionality reduction processing to obtain a first reduced dimensionality feature; and the second simplified feature is subjected to dimensionality reduction processing to obtain a second reduced dimensionality feature.

[0032] S105. Merge the first dimensionality reduction feature and the second dimensionality reduction feature to obtain a merged feature, and input the merged feature into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature.

[0033] Among them, the dimensionality difference between the first dimensionality reduction feature and the second dimensionality reduction feature is smaller than a preset range, and the fault diagnosis model can pay attention to the first dimensionality reduction feature and the second dimensionality reduction feature at the same time, and perform fault diagnosis by combining the first dimensionality reduction feature and the second dimensionality reduction feature.

[0034] It should be noted that the dimension difference between the first dimensionality reduction feature and the second dimensionality reduction feature is smaller than a preset range.

[0035] Specifically, the first dimension reduction feature and the second dimension reduction feature are combined to obtain a combined feature. It can be understood that by combining these two types of features, the combined feature obtained can more comprehensively describe the working condition and state of the turbine cooler.

[0036] In specific implementation, the first dimensionality reduction feature and the second dimensionality reduction feature may be concatenated to obtain a merged feature.

[0037] It should be noted that when reducing the dimension of the first simplified feature and the second simplified feature, it is necessary to ensure that the dimension difference between the first reduced dimension feature and the second reduced dimension feature is less than a preset range. In this way, merging the data whose two dimensional differences are less than the preset range can avoid the situation where the fault diagnosis model only focuses on the data with high dimension and ignores the data with low dimension when one of the dimensions is too high or too low.

[0038] Furthermore, the specific range value of the preset range is set according to actual needs, and is not limited in this embodiment. In specific implementation, the preset range can be determined according to the processing performance of the fault diagnosis model.

[0039] Furthermore, the pre-trained fault diagnosis model can characterize the correlation between different states of the turbine cooler under different operating conditions and the fault conditions of the turbine cooler.

[0040] In specific implementation, the fault detection model processes the input merged features, while paying attention to the first dimensionality reduction features that characterize the real-time operating data of the turbine cooler and the second dimensionality reduction features that characterize the real-time status information, and determines the current fault condition by combining the correlation between different states under different operating conditions and the fault condition of the turbine cooler.

[0041] A specific embodiment will be given below to explain in detail the training process of the fault detection model, which will not be repeated here.

[0042] The fault diagnosis method for a turbine cooler provided in the present application obtains real-time operating data and real-time status signals of the turbine cooler, and performs feature extraction on the real-time operating data and the real-time status signals to obtain real-time operating characteristics corresponding to the real-time operating data, and real-time status characteristics corresponding to the real-time status signals, and then simplifies the real-time operating characteristics and the real-time status characteristics respectively to obtain a first simplified feature corresponding to the real-time operating characteristics, and a second simplified feature corresponding to the real-time status characteristics, and performs dimensionality reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimensionality reduction feature corresponding to the first simplified feature, and a second dimensionality reduction feature corresponding to the second simplified feature, and then merges the first dimensionality reduction feature and the second dimensionality reduction feature to obtain a merged feature, and inputs the merged feature into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature. In this way, on the first hand, by performing feature extraction, simplification processing and dimensionality reduction processing on the real-time working condition data and the real-time state signal respectively, the dimension difference between the dimensionality reduction features corresponding to the working condition data and the dimensionality reduction features corresponding to the state signal is smaller than the preset range. In this way, not only can the working condition data and the state information be combined for fault diagnosis at the same time, but also the negotiation between the two in the feature dimension can be ensured, ensuring that the fault diagnosis model can simultaneously and balancedly consider the working condition characteristics and the state characteristics in the process of fault diagnosis, and combine the working condition characteristics and the state characteristics to perform fault diagnosis, so as to accurately capture the fault characteristics related to the working condition changes, and make a comprehensive judgment in combination with the state characteristics. In this way, the accuracy and reliability of fault diagnosis can be improved, and the adaptability of the fault diagnosis model to different working conditions can be enhanced; on the second hand, through feature simplification and dimensionality reduction, the dimension of the original feature data is reduced, the redundant information and the computational complexity are reduced, so that the dimension difference between the final first dimensionality reduction feature and the second dimensionality reduction feature is smaller than the preset range, which helps the fault diagnosis model to effectively focus on key features and avoid the model overfitting problem caused by high-dimensional data.

[0043] Figure 2 This is a flow chart of Embodiment 2 of the fault diagnosis method for a turbine cooler provided in this application. Figure 2 Based on the above embodiment, the training process of the pre-trained fault diagnosis model may include: S201. Acquire multiple groups of test data of a turbine cooler in a normal state and a fault state; wherein each group of test data includes operating condition data and a status signal.

[0044] Specifically, the normal state refers to a state in which all parts of the turbo cooler are intact. It is understandable that in the normal state, the turbo cooler operates under various working conditions, and its various parameters are within the normal range and can operate stably.

[0045] Furthermore, the fault state refers to the state of the turbo cooler when a fault occurs. It is understandable that in the fault state, the turbo cooler has parameters that are not within the normal range when it is running and cannot operate stably.

[0046] In specific implementation, the turbine cooler is operated in a normal state and a fault state, and the operating condition data and status signals thereof are measured and recorded by sensors. For example, in one embodiment, a normal test and a fault injection test of the turbine cooler can be carried out, and test data can be obtained.

[0047] It should be noted that the fault types included in the fault condition are set according to actual needs and are not limited in this embodiment. For example, in a possible implementation, the fault types included in the fault condition may include: unbalance fault, bearing fault, and blade crack fault.

[0048] S202. For each group of test data, slice the group of test data according to preset rules to obtain multiple slice data of the group of test data, and perform feature extraction on the operating condition data and state signal in each slice data to obtain the historical operating condition features corresponding to the operating condition data and the historical state features corresponding to the state signal.

[0049] In specific implementation, after obtaining multiple groups of test data, feature extraction is performed on the operating condition data and state signals in each group of test data to obtain historical operating condition features and historical state features.

[0050] Optionally, in a possible implementation, the operating condition data may include multiple specified operating condition parameters of the following operating condition parameters: rotation speed, inlet temperature of the fan side, outlet temperature of the fan side, inlet temperature of the turbine side, outlet temperature of the turbine side, inlet flow of the fan side, outlet flow of the fan side, inlet flow of the turbine side, outlet flow of the turbine side, inlet pressure of the fan side, outlet pressure of the fan side, inlet pressure of the turbine side, and outlet pressure of the turbine side; further, at this time, the process of extracting features from the operating condition data may include: For any specified operating condition parameter, the mean value of the operating condition parameter is determined, and the mean values ​​of all specified operating condition parameters are determined as the operating condition characteristics corresponding to the operating condition data.

[0051] In specific implementation, the specified operating condition parameters may include all the above operating condition parameters, or some of the above operating condition parameters, which are not limited in this embodiment. The following description is made by taking the example that the specified operating condition parameters include all the above operating condition parameters.

[0052] Furthermore, when performing feature extraction, all specified operating condition parameters may be traversed, and the mean of the operating condition parameter currently traversed may be calculated, and then the mean of all specified operating condition parameters may be determined as the operating condition feature corresponding to the operating condition data.

[0053] It should be noted that when extracting features from operating condition data, the average operating state of the turbine cooler under different operating conditions can be captured by calculating the mean of the specified operating condition parameters. In this way, the operating condition features corresponding to the operating condition data can be accurately obtained.

[0054] Furthermore, in a possible implementation, the process of extracting features from the state signal may include: (1) A signal segment of a preset time length is taken from the state signal, and a time domain feature of the signal segment in the time domain is determined; wherein the time domain feature includes at least an effective value and a peak-to-peak value.

[0055] Specifically, the specific duration of the preset duration is set according to actual needs, and is not limited in this embodiment. For example, in one embodiment, the preset duration can be set to 1 second.

[0056] Furthermore, the effective value is the time average of the amplitude of the signal segment, and the energy or power level of the signal segment can be described by the effective value.

[0057] Furthermore, the peak-to-peak value is the difference between the maximum and minimum values ​​in the waveform of a signal segment. The peak-to-peak value can reflect the dynamic range and fluctuation amplitude of the signal segment.

[0058] In specific implementation, after extracting a signal segment of a preset length from the state signal, the square of the amplitude of the signal segment is time-averaged and then squared to obtain the effective value; further, the maximum and minimum values ​​in the signal segment are determined, and the difference between the two is calculated to obtain the peak-to-peak value.

[0059] (2) Performing Fourier transform on the signal segment to obtain a frequency spectrum corresponding to the signal segment, and extracting a first type of frequency domain feature based on the frequency spectrum; wherein the first type of frequency domain feature at least includes an amplitude of N times the fundamental frequency; and N is a positive integer.

[0060] Specifically, the spectrum corresponding to the signal segment is the representation of the signal segment in the frequency domain after Fourier transformation. It can be understood that the spectrum can display the amplitude and phase information of each frequency component in the signal segment, and the frequency components and their intensities of the signal segment can be observed on the spectrum.

[0061] Furthermore, the specific data type included in the first type of frequency domain features is determined according to actual needs, and is not limited in this embodiment. In specific implementation, the amplitude of the N-fold frequency of the base frequency can be determined as the first type of frequency domain features.

[0062] It should be noted that the specific data of N is set according to actual needs and is not limited in this embodiment. For example, N may include 1, 2, and 3. In this case, the amplitude of the fundamental frequency, the amplitude of the double frequency of the fundamental frequency, and the amplitude of the triple frequency of the fundamental frequency are determined as the first type of frequency domain features.

[0063] (3) For each of the preset division rules, the spectrum is divided into a plurality of sub-spectra according to the division rule, and for the spectrum and at least one of the plurality of sub-spectra, a second type of frequency domain feature of the spectrum is extracted; wherein the second type of frequency domain feature includes at least a centroid frequency, a mean square frequency, and a frequency variance.

[0064] Specifically, the preset division rule is used to divide the spectrum, and the spectrum is divided according to the preset division rule to obtain a plurality of sub-spectra.

[0065] It should be noted that the specific number of preset division rules and the specific content of each preset division rule are set according to actual needs, and are not limited in this embodiment. In specific implementation, for example, in a possible implementation, the preset division rule can be determined according to the characteristics of the spectrum. For example, in one embodiment, a sub-band can be divided every 100 Hz.

[0066] In a possible implementation, the preset division rule includes at least one of a first rule and a second rule; wherein the first rule is to divide the spectrum according to a preset frequency interval; and the second rule is to divide the spectrum with the plunger passing frequency and the multiples of the plunger passing frequency as the center of each sub-spectrum, and half of the plunger passing frequency as the left neighboring interval and the right neighboring interval of the sub-spectrum.

[0067] Furthermore, for each division rule, after obtaining a plurality of sub-spectra based on the division rule, the second type of frequency domain features of the spectrum can be further extracted for the spectrum and at least one of the plurality of sub-spectra.

[0068] It should be noted that, when extracting the second type of frequency domain features, the number of specific spectra targeted is set according to actual needs, and is not limited in this embodiment. For example, in a possible implementation, the second type of frequency domain features can be extracted for the spectrum and all spectra in the multiple sub-spectra. The second type of frequency domain features can also be extracted for the spectrum and some sub-spectra in the multiple sub-spectra.

[0069] The specific data type included in the second type of frequency domain features is determined according to actual needs, and is not limited in this embodiment. In specific implementation, the second frequency domain may include centroid frequency, mean square frequency, and frequency variance.

[0070] For a certain spectrum, the specific calculation principle of the centroid frequency, mean square frequency and frequency variance can be referred to the description in the relevant technology, which will not be repeated here.

[0071] (4) Decomposing the signal segment into a plurality of sub-frequency bands using wavelet packet decomposition technology according to a preset number of decomposition levels; wherein at least one number of decomposition levels is preset.

[0072] Specifically, the specific value of the preset number of decomposition layers is set according to actual needs, and is not limited in this embodiment. For example, in one embodiment, the preset number of decomposition layers is 3 layers; for another example, in another embodiment, the preset number of decomposition layers is 5 layers.

[0073] It should be noted that the wavelet packet decomposition technology is used to perform fine signal analysis on signal segments. Wavelet packet decomposition technology provides higher frequency resolution than traditional wavelet decomposition, and can perform multi-level and multi-scale decomposition of signals, decomposing signal segments into multiple sub-bands while retaining the local time-frequency characteristics of the signal segments. The specific implementation process and implementation principle of the wavelet packet decomposition technology can be found in the description of the relevant technology, which will not be repeated here.

[0074] (5) For each sub-band, determine the energy proportion of the sub-band, and determine the energy proportion of all sub-bands as the third type of frequency domain features.

[0075] In specific implementation, the energy of each sub-band is divided by the total energy to obtain the energy proportion of each sub-band, and the energy proportion of all sub-bands is determined as the third type of frequency domain feature.

[0076] (6) Determine the time domain feature, the first type of frequency domain feature, the second type of frequency domain feature, and the third type of frequency domain feature as the state feature of the state signal.

[0077] In specific implementation, the time domain features, the first type of frequency domain features, the second type of frequency domain features, and the third type of frequency domain features may be spliced ​​and determined as the state features corresponding to the state signal in the group of test data.

[0078] As described above, when extracting features from the state signal, the amplitude of the fundamental frequency and its multiples in the first frequency domain feature can reflect the periodic components of the signal, identify the inherent frequency and harmonic components of the system, and help determine the operating status and potential faults of the turbine cooler. The center of gravity frequency in the second frequency domain feature reflects the concentrated position of the signal energy, which helps identify the main frequency components of the signal. Furthermore, the mean square frequency and frequency variance provide statistical characteristics of the signal frequency distribution, which helps analyze the stability and complexity of the signal. At the same time, the wavelet packet decomposition technology can provide fine structural information of the signal at different frequency scales. By calculating the energy share of each sub-band, the energy distribution of the signal in different frequency ranges can be revealed. In this way, the feature extraction of the state signal can be completed accurately.

[0079] It can be understood that when extracting features from the status signal, the time domain features, the first type of frequency domain features, the second type of frequency domain features, and the third type of frequency domain features are comprehensively considered. These features comprehensively reflect the status characteristics of the turbine cooler, avoid the problem of loss of key features caused by incomplete features, and can further improve the accuracy of diagnosis.

[0080] It should be noted that, referring to the previous description, it can be seen that after feature extraction of the operating condition data and the state signal, the number of state features is much greater than the number of operating condition features. For example, in one embodiment, the number of state features is 80 and the number of operating condition features is 9.

[0081] In addition, it can be understood that the feature extraction process of the operating condition data and the status signal is similar during the training phase of the fault detection model and the use phase of the fault detection model. In other words, in Example 1, the feature extraction of the operating condition data and the status signal can also be performed according to the method provided in this step.

[0082] S203 , simplify the historical operating condition feature and the historical state feature respectively to obtain a first historical simplified feature corresponding to the operating condition feature and a second historical simplified feature corresponding to the state feature.

[0083] In specific implementation, the historical operating condition characteristics are simplified to obtain the first historical simplified characteristics; the historical state characteristics are simplified to obtain the second historical simplified characteristics.

[0084] Optionally, in a possible implementation, the specific process of the simplification process may include: (1) For a feature to be simplified, the weight of each sub-feature in the feature is determined based on the relief algorithm; the feature to be simplified is a working condition feature or a state feature.

[0085] Furthermore, the features to be simplified are working condition features and state features. The following is an example of state features. In specific implementation, the weight of each sub-feature in the working condition feature is determined by the relief algorithm. It can be understood that the weight of a sub-feature represents the contribution of the sub-feature to distinguishing the adjacent sub-feature categories. In other words, the greater the weight of the sub-feature, the greater the contribution of the sub-feature to distinguishing the adjacent sub-features.

[0086] (2) Deleting the target sub-feature of the feature; wherein the target sub-feature is a sub-feature whose weight is less than or equal to 0.

[0087] Figure 3 This is a schematic diagram of the weights of each sub-feature in the state feature shown in an exemplary embodiment of the present application. Figure 3 , the larger the weight, the more effective the sub-feature is for fault identification, and when the weight is less than 0, it means that the sub-feature has a negative effect on fault identification. Therefore, in this step, sub-features with weights less than or equal to 0 are eliminated.

[0088] It is understandable that when simplifying features, the importance of sub-features is evaluated through the relieff algorithm, and target sub-features with weights less than or equal to 0 are deleted, which can effectively reduce the number of sub-features and reduce the dimension of the data. In this way, the simplification process can prevent the model from overfitting during operation. At the same time, by reducing the data dimensions of the operating data and the state signal to the same level through simplification, it can be ensured that the operating data with a low amount of original data will not be ignored in subsequent analysis and calculation. In this way, the model can fully consider the effect of the operating data on the turbine cooler and improve the reliability and accuracy of fault detection.

[0089] It should be noted that the implementation process of simplifying features is consistent during the model training phase and the model usage phase. In other words, in Example 1, features can be simplified in a manner similar to this step.

[0090] S204: Perform dimensionality reduction processing on the first historical simplified feature and the second historical simplified feature respectively to obtain a first historical dimensionality reduction feature corresponding to the first historical simplified feature and a second historical dimensionality reduction feature corresponding to the second historical simplified feature.

[0091] In a specific implementation, the first historical simplified feature is subjected to dimensionality reduction processing to obtain the first historical reduced dimensionality feature; and the second historical simplified feature is subjected to dimensionality reduction processing to obtain the second historical reduced dimensionality feature.

[0092] Optionally, in a possible implementation, when performing dimensionality reduction processing on simplified features, a principal component analysis method may be used to perform dimensionality reduction processing on simplified features.

[0093] When implementing it specifically, Figure 4 This is a schematic diagram showing the information ratio of each component included in a simplified feature of an exemplary embodiment of the present application. Figure 4 , based on the principal component analysis method to simplify the features, set the parameters to retain more than 90% of the data, and the retained components are the data after dimensionality reduction. Figure 4 It can be seen that the data after dimensionality reduction contains more than 90% of the information of the original data.

[0094] It should be noted that the specific implementation process of dimensionality reduction is consistent during the model training phase and the model usage phase. In other words, in Example 1, dimensionality reduction can be performed on simplified features based on the method provided in this step.

[0095] Referring to the above description, it can be understood that after simplifying and reducing the dimension of the working condition feature, and simplifying and reducing the dimension of the state feature, the dimension difference between the reduced dimension feature corresponding to the working condition feature and the reduced dimension feature corresponding to the state feature is small. For example, in one embodiment, the reduced dimension feature corresponding to the working condition feature is 3, and the reduced dimension feature corresponding to the state feature is 5.

[0096] S205: Merge the first historical dimension reduction feature and the second historical dimension reduction feature to obtain a historical merged feature corresponding to each slice data.

[0097] It should be noted that the first historical dimension reduction feature and the second historical dimension reduction feature can be spliced ​​together to obtain a historical merged feature.

[0098] S206. Take the historical merged features corresponding to each slice data of each group of test data as a sample, and take the encoding information of the state corresponding to the group of test data as the label of the sample, so as to construct a training sample; wherein, multiple historical merged features corresponding to multiple slice data of each group of test data constitute multiple training samples.

[0099] Specifically, the coding information of the health status of the turbine cooler is used to distinguish different states of the turbine cooler. In specific implementation, the normal state, unbalance fault, bearing fault, blade crack fault and other states of the turbine cooler can be coded respectively. For example, in one embodiment, the coding information of the normal state, unbalance fault, bearing fault and blade crack fault is determined as 1, 2, 3 and 4 respectively.

[0100] In the specific implementation, the encoding information of the health status of the turbine cooler corresponding to each set of test data is used as the label of the historical merge feature to construct the training sample. It should be noted that using the encoding information as the label of the sample can help the model distinguish different health status of the turbine cooler during training.

[0101] S207: Use the training samples to train an initial model to obtain the pre-trained fault diagnosis model.

[0102] Specifically, the initial model refers to a model whose structure has been defined but has not yet been trained with data before training.

[0103] Optionally, in a possible implementation, the initial model is any one of the following models: support vector machine, K nearest neighbor, multi-layer perceptron, decision tree.

[0104] In the specific implementation, the process of training the initial model based on the training samples can be found in the description of the relevant technology and will not be repeated here.

[0105] Furthermore, in order to verify the accuracy of the method provided by this implementation, this paper also verifies the above method. Figure 5 A schematic diagram of the confusion matrix of the fault diagnosis model provided in this application. Please refer to Figure 5 , Figure 5 The value at the ijth position in represents the number of samples of category i that are predicted to be category j. Please refer to Figure 5 ,The error rate of the fault diagnosis model is controlled within 2%, and the accuracy is good.

[0106] The fault diagnosis method for a turbine cooler provided in the present embodiment obtains multiple groups of test data of the turbine cooler under normal working conditions and faulty working conditions, and then for each group of test data, extracts features of the working condition data and the state signal in the group of test data respectively, obtains the historical working condition features corresponding to the working condition data, and the historical state features corresponding to the state signal, and simplifies the historical working condition features and the historical state features respectively to obtain the first historical simplified features corresponding to the working condition features, and the second historical simplified features corresponding to the state features, and then performs dimensionality reduction processing on the first historical simplified features and the second historical simplified features respectively to obtain the first historical reduced dimensionality features corresponding to the first historical simplified features, and the second historical reduced dimensionality features corresponding to the second historical simplified features, and finally merges the first historical reduced dimensionality features and the second historical reduced dimensionality features to obtain the historical merged features corresponding to the group of test data, and takes the historical merged features corresponding to each group of test data as a sample, and takes the encoding information of the working condition corresponding to the group of test data as the label of the sample, so as to construct a training sample, thereby using the training sample to train the initial model to obtain the pre-trained fault diagnosis model. In this way, a fault diagnosis model can be trained and then used to accurately diagnose faults of the turbine cooler. In addition, when training the model, by extracting features from multiple groups of test data under normal and fault conditions and simplifying and reducing the dimensions of the features, not only can redundancy and noise be reduced, but also the difference in the dimensions of the reduced features of the operating data after dimensionality reduction and the reduced features of the state signal after dimensionality reduction can be ensured to be within a preset range, which can increase the model's attention to both and ensure that the model can focus on the most diagnostically significant features when dealing with complex situations in practical applications to improve accuracy. Furthermore, feature simplification and dimensionality reduction not only reduce the dimension of the data, but also reduce the complexity of calculation, reduce the calculation time in the training and reasoning process, and improve the operating efficiency of the model. Moreover, the reduced-dimensional data enables the model to learn more efficiently and avoid overfitting problems caused by high-dimensional data.

[0107] Corresponding to the aforementioned embodiment of a fault diagnosis method for a turbine cooler, the present application also provides an embodiment of a fault diagnosis device for a turbine cooler.

[0108] The present application provides an embodiment of a fault diagnosis device for a turbine cooler that can be applied to a fault diagnosis device for a turbine cooler. The device embodiment can be implemented through software, or through hardware, or a combination of software and hardware. Taking software implementation as an example, as a device in a logical sense, the processor of the fault diagnosis device for the turbine cooler in which it is located reads the corresponding computer program instructions in the non-volatile memory into the memory and runs them. From a hardware perspective, if Figure 6As shown, it is a hardware structure diagram of the fault diagnosis device of the turbine cooler where the fault diagnosis device of the turbine cooler of the present application is located, except Figure 6 In addition to the processor, memory, network interface, and non-volatile memory shown, the fault diagnosis device of the turbine cooler where the device in the embodiment is located may also include other hardware according to the actual function of the fault diagnosis device of the turbine cooler, which will not be described in detail.

[0109] Figure 7 This is a schematic diagram of the structure of the first embodiment of the fault diagnosis device for the turbine cooler provided in this application. Figure 7 The device provided in this embodiment includes an acquisition module 710, an extraction module 720, a simplification module 730, a dimension reduction module 740 and a processing module 750; wherein: The acquisition module 710 is used to acquire real-time operating data and real-time status signals of the turbine cooler; The extraction module 720 is used to extract features from the real-time operating condition data and the real-time status signal respectively, to obtain the real-time operating condition features corresponding to the real-time operating condition data and the real-time status features corresponding to the real-time status signal; The simplification module 730 is used to simplify the real-time working condition feature and the real-time state feature respectively to obtain a first simplified feature corresponding to the real-time working condition feature and a second simplified feature corresponding to the real-time state feature; The dimension reduction module 740 is used to perform dimension reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimension reduction feature corresponding to the first simplified feature and a second dimension reduction feature corresponding to the second simplified feature; The processing module 750 is used to merge the first dimensionality reduction feature and the second dimensionality reduction feature to obtain a merged feature, and input the merged feature into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature; wherein the dimensional difference between the first dimensionality reduction feature and the second dimensionality reduction feature is less than a preset range, and the fault diagnosis model can pay attention to the first dimensionality reduction feature and the second dimensionality reduction feature at the same time, and perform fault diagnosis by combining the first dimensionality reduction feature and the second dimensionality reduction feature.

[0110] The device of this embodiment can be used to perform Figure 1 The steps, specific implementation principles and implementation processes of the method embodiment shown are similar and will not be repeated here.

[0111] Optionally, the extraction module 720 is specifically used to extract a signal segment of a preset duration from the state signal, and determine a time domain feature of the signal segment in the time domain; wherein the time domain feature includes at least an effective value and a peak-to-peak value; The extraction module 720 is further specifically configured to perform Fourier transform on the signal segment to obtain a frequency spectrum corresponding to the signal segment, and extract a first type of frequency domain feature according to the frequency spectrum; wherein the first type of frequency domain feature at least includes an amplitude of N times the fundamental frequency; and N is a positive integer; The extraction module 720 is further specifically configured to divide the spectrum into a plurality of sub-spectra according to each of the preset division rules, and extract a second type of frequency domain feature of the spectrum for the spectrum and at least one of the plurality of sub-spectra; wherein the second type of frequency domain feature includes at least a centroid frequency, a mean square frequency, and a frequency variance; The extraction module 720 is further specifically configured to decompose the signal segment into a plurality of sub-frequency bands using a wavelet packet decomposition technique according to a preset number of decomposition levels; wherein at least one number of decomposition levels is preset; The extraction module 720 is further specifically configured to determine the energy proportion of each sub-frequency band, and determine the energy proportions of all sub-frequency bands as the third type of frequency domain features; The extraction module 720 is further specifically configured to determine the time domain feature, the first type of frequency domain feature, the second type of frequency domain feature, and the third type of frequency domain feature as the state feature of the state signal.

[0112] Optionally, the extraction module 720 is further specifically used to determine the mean value of any specified operating condition parameter, and determine the mean values ​​of all specified operating condition parameters as the operating condition feature corresponding to the operating condition data.

[0113] Optionally, the simplification module 730 is specifically used to determine the weight of each sub-feature in the feature to be simplified based on the relief algorithm; the feature to be simplified is a working condition feature or a state feature; The reduction module 730 is further specifically configured to delete a target sub-feature of the feature; wherein the target sub-feature is a sub-feature whose weight is less than or equal to 0.

[0114] Optionally, the dimension reduction module 740 is specifically used to perform dimension reduction processing on simplified features using a principal component analysis method.

[0115] Optionally, the preset division rule includes at least one of a first rule and a second rule; wherein the first rule is to divide the spectrum according to a preset frequency interval; and the second rule is to divide the spectrum with the plunger passing frequency and the multiples of the plunger passing frequency as the center of each sub-spectrum, and half of the plunger passing frequency as the left neighboring interval and the right neighboring interval of the sub-spectrum.

[0116] Optionally, the initial model is any one of the following models: support vector machine, K nearest neighbor, multi-layer perceptron, decision tree.

[0117] Please continue to refer to Figure 6 The present application also provides a fault diagnosis device for a turbine cooler, comprising 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 steps of any one of the methods provided in the first aspect of the present application are implemented.

[0118] The present application also provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the steps of any one of the methods provided in the present application are implemented.

[0119] The implementation process of the functions and effects of each unit in the above-mentioned device is specifically described in the implementation process of the corresponding steps in the above-mentioned method, and will not be repeated here.

[0120] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0121] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A fault diagnosis method for a turbine cooler, characterized in that: The method comprises: Obtain real-time operating data and real-time status signals of the turbine cooler; Performing feature extraction on the real-time operating condition data and the real-time status signal respectively to obtain real-time operating condition features corresponding to the real-time operating condition data and real-time status features corresponding to the real-time status signal; Simplifying the real-time working condition feature and the real-time state feature respectively to obtain a first simplified feature corresponding to the real-time working condition feature and a second simplified feature corresponding to the real-time state feature; Performing dimensionality reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimensionality reduction feature corresponding to the first simplified feature and a second dimensionality reduction feature corresponding to the second simplified feature; The first dimensionality reduction feature and the second dimensionality reduction feature are merged to obtain a merged feature, and the merged feature is input into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature; wherein the dimensionality difference between the first dimensionality reduction feature and the second dimensionality reduction feature is less than a preset range, and the fault diagnosis model can pay attention to the first dimensionality reduction feature and the second dimensionality reduction feature at the same time, and perform fault diagnosis by combining the first dimensionality reduction feature and the second dimensionality reduction feature.

2. The method according to claim 1, characterized in that The training process of the pre-trained fault diagnosis model includes: Acquire multiple groups of test data of the turbine cooler in normal state and fault state; wherein each group of test data includes operating condition data and status signal; For each group of test data, the group of test data is sliced ​​according to a preset rule to obtain multiple slice data of the group of test data, and feature extraction is performed on the working condition data and the state signal in each slice data to obtain the historical working condition feature corresponding to the working condition data and the historical state feature corresponding to the state signal; Simplifying the historical operating condition feature and the historical state feature respectively to obtain a first historical simplified feature corresponding to the operating condition feature and a second historical simplified feature corresponding to the state feature; Performing dimensionality reduction processing on the first historical simplified feature and the second historical simplified feature respectively to obtain a first historical dimensionality reduction feature corresponding to the first historical simplified feature and a second historical dimensionality reduction feature corresponding to the second historical simplified feature; Merging the first historical dimension reduction feature and the second historical dimension reduction feature to obtain a historical merged feature corresponding to each slice data; The historical merged features corresponding to each slice data of each group of test data are taken as a sample, and the encoding information of the state corresponding to the group of test data is taken as the label of the sample, so as to construct a training sample; wherein, the multiple historical merged features corresponding to the multiple slice data of each group of test data constitute multiple training samples; The training samples are used to train an initial model to obtain the pre-trained fault diagnosis model.

3. The method according to claim 1 or 2, characterized in that: The process of feature extraction of state signals includes: Taking a signal segment of a preset time length from the state signal, and determining the time domain characteristics of the signal segment in the time domain; wherein the time domain characteristics at least include an effective value and a peak-to-peak value; Performing Fourier transform on the signal segment to obtain a frequency spectrum corresponding to the signal segment, and extracting a first type of frequency domain feature according to the frequency spectrum; wherein the first type of frequency domain feature at least includes an amplitude of N times the fundamental frequency; N is a positive integer; For each division rule in the preset division rules, the spectrum is divided into a plurality of sub-spectra according to the division rule, and for the spectrum and at least one spectrum in the plurality of sub-spectra, a second type of frequency domain feature of the spectrum is extracted; wherein the second type of frequency domain feature includes at least a centroid frequency, a mean square frequency and a frequency variance; For a preset number of decomposition levels, the signal segment is decomposed into a plurality of sub-frequency bands using a wavelet packet decomposition technique; wherein at least one number of decomposition levels is preset; For each sub-band, determining the energy proportion of the sub-band, and determining the energy proportions of all sub-bands as the third type of frequency domain features; The time domain feature, the first type of frequency domain feature, the second type of frequency domain feature, and the third type of frequency domain feature are determined as the state features of the state signal.

4. The method according to claim 1 or 2, characterized in that: The operating condition data includes a plurality of specified operating condition parameters among the following operating condition parameters: rotation speed, inlet temperature of the fan side, outlet temperature of the fan side, inlet temperature of the turbine side, outlet temperature of the turbine side, inlet flow rate of the fan side, outlet flow rate of the fan side, inlet flow rate of the turbine side, outlet flow rate of the turbine side, inlet pressure of the fan side, outlet pressure of the fan side, inlet pressure of the turbine side, and outlet pressure of the turbine side; The process of extracting features from working condition data includes: For any specified operating condition parameter, the mean value of the operating condition parameter is determined, and the mean values ​​of all specified operating condition parameters are determined as the operating condition characteristics corresponding to the operating condition data.

5. The method according to claim 1 or 2, characterized in that: The process of simplifying the working condition characteristics and state characteristics respectively includes: For a feature to be simplified, the weight of each sub-feature in the feature is determined based on a relief algorithm; the feature to be simplified is a working condition feature or a state feature; Delete the target sub-feature of the feature; wherein the target sub-feature is a sub-feature whose weight is less than or equal to 0.

6. The method according to claim 1 or 2, characterized in that: The process of reducing the dimensionality of simplified features includes: The principal component analysis method is used to reduce the dimension of simplified features.

7. The method according to claim 3, characterized in that The preset division rule includes at least one of a first rule and a second rule; wherein the first rule is to divide the spectrum according to a preset frequency interval; and the second rule is to divide the spectrum with the plunger passing frequency and the multiples of the plunger passing frequency as the center of each sub-spectrum, and half of the plunger passing frequency as the left neighboring interval and the right neighboring interval of the sub-spectrum.

8. The method according to claim 2, characterized in that: The initial model is any one of the following models: support vector machine, K nearest neighbor, multi-layer perceptron, decision tree.

9. A fault diagnosis device for a turbine cooler, characterized in that: The device comprises an acquisition module, an extraction module, a simplification module, a dimension reduction module and a processing module; wherein: The acquisition module is used to acquire real-time operating condition data and real-time status signals of the turbine cooler; The extraction module is used to extract features from the real-time operating condition data and the real-time status signal respectively, to obtain the real-time operating condition features corresponding to the real-time operating condition data and the real-time status features corresponding to the real-time status signal; The simplification module is used to simplify the real-time working condition feature and the real-time state feature respectively to obtain a first simplified feature corresponding to the real-time working condition feature and a second simplified feature corresponding to the real-time state feature; The dimension reduction module is used to perform dimension reduction processing on the first simplified feature and the second simplified feature respectively to obtain a first dimension reduction feature corresponding to the first simplified feature and a second dimension reduction feature corresponding to the second simplified feature; The processing module is used to merge the first dimensionality reduction feature and the second dimensionality reduction feature to obtain a merged feature, and input the merged feature into a pre-trained fault diagnosis model so that the fault diagnosis model can identify the fault of the turbine cooler based on the merged feature; wherein the dimensional difference between the first dimensionality reduction feature and the second dimensionality reduction feature is less than a preset range, and the fault diagnosis model can pay attention to the first dimensionality reduction feature and the second dimensionality reduction feature at the same time, and perform fault diagnosis by combining the first dimensionality reduction feature and the second dimensionality reduction feature.

10. A fault diagnosis device for a turbine cooler, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 8 when executing the program.

Images