A method and device for online diagnosis of turbine cooler

Abstract

The present application provides a method and device for online diagnosis of a turbine cooler, which belongs to the field of aviation electromechanics. The method provided by the present application includes: determining the type of monitoring signal and the monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is the signal with the largest amplitude change under the fault state; real-time acquisition of the monitoring signal; independent and separate identification of whether the turbine cooler is faulty based on any one of the monitoring signals; if the identification result is no fault, fusing each signal in the monitoring signal, inputting the fused signal into a neural network model, and outputting the fault state of the turbine cooler. The method and device for online diagnosis of a turbine cooler provided by the present application can realize fast and accurate fault diagnosis under complex working conditions.

Description

Technical Field

[0001] The present application relates to the field of aviation electromechanical technology, and in particular to an online diagnosis method and device for a turbine cooler. Background Art

[0002] As a high-speed rotating machine, the turbine cooler often suffers from malfunctions such as jamming, breaking, and abnormal wear during testing and use. Once these malfunctions occur, they will not only lead to serious consequences, but may also affect flight safety. Therefore, early detection of product operation abnormalities and fault diagnosis are of great significance to improving product quality.

[0003] At present, aircraft turbine coolers are still mainly pure mechanical products, with relatively scarce monitoring methods, and mainly rely on regular maintenance. This maintenance method easily leads to the failure to be discovered in time, which in turn leads to the spread of the failure, or the increase of costs and disassembly risks due to frequent inspections. Therefore, it is necessary to monitor the operating status of the turbine cooler in real time online to detect faults in time and issue early warnings.

[0004] However, due to the aerodynamic influence and large fluctuation of the operating speed of the turbine cooler during operation, the identification of fault characteristic signals is seriously affected. At present, the online fault diagnosis of the turbine cooler cannot achieve accurate fault location under complex working conditions, and the fault identification efficiency is low due to the influence of multiple fault characteristic signals. Summary of the invention

[0005] In view of this, the present application provides a turbine cooler online diagnosis method and device, which can achieve fast and accurate fault diagnosis under complex working conditions.

[0006] Specifically, the present application is implemented through the following technical solutions:

[0007] A first aspect of the present application provides a turbine cooler online diagnosis method, the method comprising:

[0008] Determine the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is the signal with the largest amplitude change under the fault state;

[0009] Collecting the monitoring signal in real time;

[0010] independently and individually identifying whether the turbine cooler is faulty based on any one of the monitoring signals;

[0011] Wherein, the recognition order of each signal is determined according to the time sequence of the associated faults of each signal in the monitoring signal;

[0012] If the identification result is that there is no fault, the various signals in the monitoring signal are fused, the fused signal is input into the neural network model, and the fault state of the turbine cooler is output;

[0013] Wherein, numerical parameters of at least two dimensions are calculated based on each signal, and each signal is fused based on the fusion of the numerical parameters.

[0014] A second aspect of the present application provides an online diagnostic device for a turbine cooler, the device comprising a determination module, a collection module, an identification module and a processing module;

[0015] wherein the determination module is used to determine the monitoring signal type and the monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is the signal with the largest amplitude change under the fault state;

[0016] The acquisition module is used to acquire the monitoring signal in real time;

[0017] The identification module is used to independently identify whether the turbine cooler is faulty based on any one of the monitoring signals;

[0018] Wherein, the recognition order of each signal is determined according to the time sequence of the associated faults of each signal in the monitoring signal;

[0019] The processing module is used to fuse the various signals in the monitoring signal when the recognition result is no fault, input the fused signal into the neural network model, and output the fault state of the turbine cooler;

[0020] Wherein, numerical parameters of at least two dimensions are calculated based on each signal, and each signal is fused based on the fusion of the numerical parameters.

[0021] The turbine cooler online diagnosis method and device provided in the present application can directly determine the sensor type and layout position by determining the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, thereby avoiding the structural limitations, excessive economic costs and information redundancy caused by installing too many sensors, and avoiding the problems of insufficient health status recognition ability, inability to achieve early warning of faults and affecting system safety caused by installing too few sensors. In addition, the monitoring position is determined according to the fault mechanism, and monitoring at a suitable position can also improve the accuracy of fault diagnosis. In addition, it is preferred to independently identify whether the turbine cooler is faulty based on any one of the monitoring signals, which utilizes the different fault sensitivities of each signal in the monitoring signal for analysis, wherein the signal with the largest amplitude change under the fault state has the highest fault sensitivity. The recognition order of each signal can be obtained by using the sensitivity from high to low. Faults can be identified based on this order, which can achieve rapid fault identification and improve the efficiency of fault identification. When any signal has not yet identified the fault, the features of each signal can be extracted and fused to integrate the information of each signal, avoid interference from the external working environment of the turbine, and increase the comprehensiveness and accuracy of the diagnosis. It should also be noted that this dual mechanism of independent signal identification and signal fusion can reduce false positives or negative positives caused by single signal anomalies and improve the overall reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A flowchart of a first embodiment of the online diagnosis method for a turbine cooler provided in the present application;

[0023] Figure 2 This is a schematic diagram of the structure of a neural network model shown in Example 1 of the present application;

[0024] Figure 3 This is a schematic structural diagram of Embodiment 2 of the turbine cooler online diagnostic device provided in the present application. DETAILED DESCRIPTION

[0025] 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.

[0026] 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.

[0027] 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".

[0028] Specific embodiments are given below to introduce the technical solution of the present application in detail.

[0029] Figure 1 This is a flow chart of Example 1 of the online diagnosis method for a turbine cooler provided in this application. Please refer to Figure 1 , the method provided in this embodiment may include:

[0030] S101. Determine the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is the signal with the largest amplitude change under the fault state.

[0031] It should be noted that, based on the analysis of the failure mechanism of the turbine cooler, it can be concluded that the failure of the turbine cooler mainly comes from the bearing part and abnormal vibration. Among them, the bearing is one of the most important mechanical components in the turbine cooler, which is under the pressure of high-speed rotation and heavy load. Its temperature and vibration directly reflect the health status of the cooler. Specifically, the bearing is prone to generate heat due to friction under high-speed rotation. Abnormal temperature rise may indicate insufficient lubrication, excessive wear or other mechanical failures. Monitoring the bearing temperature can help fault diagnosis. In addition, the turbine cooler will generate inherent vibration signals during operation. Any mechanical failure (such as imbalance, looseness, breakage, etc.) will cause abnormal changes in the vibration signal. By analyzing the spectral characteristics of the vibration signal, the fault type and severity can be effectively identified. It should also be noted that the speed is one of the core parameters for measuring the working state of the turbine cooler. It can be used for both working state control and health state identification in the event of jamming, overspeed and other failures.

[0032] Therefore, based on the failure mechanism of the turbine cooler, it can be determined that the monitoring signal types include temperature, vibration, speed and other signals, and based on the failure mechanism of the turbine cooler, the monitoring position of each monitoring signal can also be determined. Specifically, taking the vibration signal as an example, when the turbine cooler fails, the machine is accompanied by abnormal vibration and noise, and its vibration signal reflects the product's fault information from the amplitude domain, frequency domain and time domain. Vibration monitoring can timely detect and identify these abnormal vibration phenomena, which is very important for evaluating the operating status of the product and diagnosing faults. However, due to the aerodynamic influence and large fluctuations in the operating speed of the turbine cooler, the vibration signal will be affected by the external vibration environment and the transmission path of the installation position itself, resulting in severe signal attenuation, resulting in unclear early fault characteristics. Therefore, determine the detection position, give priority to measuring the vibration of the fault source, and select the bearing seat position for vibration monitoring, so as to monitor the rotor imbalance and bearing wear faults of the turbine cooler to the greatest extent.

[0033] It should also be noted that each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, that is, when a fault occurs, each monitored signal will change and for each fault, at least one signal in the monitoring signal is the signal with the largest amplitude change under the fault state (this embodiment is used as an example for explanation, and in fact there may also be a weak fault stage, that is, the situation where the amplitude change is not easily observed), that is, among all the changed signals, there must be a signal with the largest amplitude change (wherein the amplitude change refers to the significant rate of change, and the requirements for the change range of different signals are inconsistent). Among them, the amplitude change under the fault state indicates the sensitivity of the signal to the fault, and the signal with the largest amplitude change indicates the signal that is most sensitive to the fault. The fault can be identified earlier and more accurately through the sensitivity of the signal. For example, in one embodiment, in the case of bearing wear, the vibration signal may show a significant change in frequency characteristics, while the change in the temperature signal may be smaller. At this time, selecting the vibration signal as the main monitoring signal can detect the bearing wear earlier. And at least one of the monitored signals is the signal with the largest amplitude change under the fault state, which helps to improve the accuracy of fault diagnosis. Specifically, if the amplitude change of a signal under the fault state is very small, it is likely to be masked by noise or other interference signals, resulting in missed detection or false detection, while the signal with the largest amplitude change helps to enhance the signal-to-noise ratio of the signal.

[0034] Specifically, by determining the monitoring signal type and monitoring location based on the fault mechanism of the turbine cooler, the sensor type and layout location can be directly determined, which not only avoids the structural limitations, high economic costs and information redundancy caused by installing too many sensors, but also avoids the problem of insufficient health status recognition ability caused by installing too few sensors, which cannot achieve early warning of faults and affect system safety. In addition, by determining the monitoring location according to the fault mechanism, monitoring at the appropriate location can also improve the accuracy of fault diagnosis. In addition, by ensuring that each signal will undergo abnormal changes when a fault occurs, the sensitivity of the monitoring system can be effectively improved, and even small faults can be captured in time, avoiding the latent development of faults. By focusing on the signal with the largest amplitude change under different fault conditions, the false alarm rate can be significantly reduced. When a fault occurs, the fault type can be quickly determined by directly identifying the signal with the largest amplitude change.

[0035] S102: Collect the monitoring signal in real time.

[0036] It should be noted that, based on the determined monitoring signal type and monitoring position, a sensor corresponding to the monitoring signal type is selected and sampling is performed using the corresponding sampling frequency according to the characteristics of each signal. Among them, the monitoring signal may specifically include a speed signal (sampled using a speed sensor), a fan end bearing temperature signal (sampled using a temperature sensor), a turbine end bearing temperature signal (sampled using a temperature sensor) and a vibration signal (sampled using a vibration sensor). The sampling frequencies of different signals are different. For example, in one embodiment, since the turbine cooler is a high-speed rotating machine, the vibration signal is sampled at a high frequency in order to obtain the fault frequency characteristics; since the temperature is a slowly varying signal, low-frequency sampling is performed in order to obtain the temperature change trend characteristics.

[0037] In addition, after the monitoring signal is acquired in real time, the acquired signal will be subjected to data verification to determine whether the signal acquired by the sensor is abnormal, which specifically includes: obtaining the historical operation data, preset threshold and time information of the monitoring signal; determining whether the acquired monitoring signal is abnormal based on the historical operation data, preset threshold and time information. Specifically, the currently acquired signal data is compared with the historical operation data to check whether there is consistency. If the difference is too large, there may be an abnormality; observe whether the fluctuation range of each signal data is within the reasonable range of the preset threshold. If a sudden change occurs, it exceeds the reasonable range of the preset threshold, and there is an abnormality; observe whether there is data at each sampling time point or whether there are repeated time points. If so, there is an abnormality.

[0038] S103, independently identifying whether the turbine cooler is faulty based on any one of the monitoring signals; wherein the identification order of each signal is determined according to the time sequence of the associated faults of each signal in the monitoring signal.

[0039] When the monitoring signals collected in real time are verified to be normal, whether the turbine cooler is faulty can be independently identified based on any one of the monitoring signals, including:

[0040] The monitoring signals are divided into first-category signals and second-category signals based on the amplitude changes of each signal in the monitoring signals under a fault state, and the signal recognition order is determined based on the time sequence of the associated faults of the first-category signals and the second-category signals; wherein the amplitude change of the first-category signals is greater than that of the second-category signals; signal processing is performed based on the first-category signals within a fixed time length and based on the signal processing results, it is determined whether the signal has a corresponding fault and the fault level; when the first-category signals do not identify a fault, signal processing is performed based on the second-category signals and based on the signal processing results, it is determined whether the signal has a corresponding fault and the fault level.

[0041] For example, in one embodiment, the first type of signal is a speed signal, and the second type of signal is a temperature signal. At this time, the speed signal will be recognized first, and then the temperature signal will be recognized. Therefore, for a fixed time length (such as 1 second), the average value and change rate of the speed signal are first calculated. When the average speed exceeds the rated maximum speed threshold, an overspeed warning is output; when the average speed is 0, a stuck warning is output; when the speed change rate exceeds the maximum fluctuation range threshold, a speed abnormality warning is output. When the speed signal is normal, the average value and change rate of the fan end bearing temperature and the turbine end bearing signal are calculated. When the fan end bearing average temperature exceeds the fan end bearing maximum temperature threshold, the fan end bearing temperature change rate exceeds the fan end bearing temperature maximum fluctuation range threshold, the turbine end bearing average temperature exceeds the turbine end bearing maximum temperature threshold, and the turbine end bearing temperature change rate exceeds the turbine end bearing temperature maximum fluctuation range threshold, an overtemperature warning is output.

[0042] Combined with the above content, any one of the monitoring signals can independently identify whether the turbine cooler is faulty. First, the speed signal directly represents the operating condition of the turbine cooler. When the speed signal is abnormal, there must be something wrong with the product, and there is no need for additional analysis through other signals. Secondly, the temperature signal. Due to the inertia of temperature signal monitoring, it can only be characterized when the fault is serious to a certain extent. Therefore, if the temperature signal is found to be abnormal at this time, the vibration signal must be abnormal and no analysis is required.

[0043] Specifically, when a fault occurs, each signal in the monitoring signal will fail. At this time, if it is necessary to identify the turbine cooler fault through a single signal, it is necessary to clarify the corresponding relationship between the signal and the fault, and then determine the identification order of each signal according to the time sequence of the associated faults of each signal in the monitoring signal. Determining the associated fault correspondence of each signal is based on any signal in the monitoring signal to independently identify whether the turbine cooler is faulty, specifically including:

[0044] Determine a target monitoring signal corresponding to each fault according to the historical fault information of the turbine cooler, wherein the target monitoring signal is a signal with the largest amplitude change under the fault condition;

[0045] Establishing a corresponding relationship based on the target monitoring signal and corresponding information of the fault;

[0046] Determine the to-be-confirmed fault corresponding to each signal in the monitoring signal based on the corresponding relationship;

[0047] Determining a first recognition order of each signal in the monitoring signal according to the occurrence time of the fault to be confirmed;

[0048] Adjusting the first recognition order according to the amplitude change rate of each signal to obtain a second recognition order;

[0049] Each signal in the monitoring signal is identified in sequence according to a second identification order, and only one signal is identified each time to determine whether the turbine cooler is faulty.

[0050] In this way, by adjusting the recognition order according to the occurrence time of the fault to be confirmed and the rate of change of the signal amplitude, possible faults can be located more quickly, and the signal with the largest rate of change is identified first, that is, the fault signal with a higher possibility is processed first, which can respond more quickly and discover potential problems faster, thereby shortening the fault diagnosis time. In addition, using historical fault data to determine the most relevant monitoring signals, and establishing a corresponding relationship between faults and signals based on these data, the type of fault that occurs can be determined based on the fault signal when a fault occurs during actual operation, thereby improving the accuracy of diagnosis.

[0051] S104. If the recognition result is no fault, fuse each signal in the monitoring signal, input the fused signal into the neural network model, and output the fault status of the turbine cooler; wherein, numerical parameters of at least two dimensions are calculated based on each signal, and each signal is fused based on the fusion of the numerical parameters.

[0052] It should be noted that if any signal independently fails to identify whether the turbine cooler is faulty, the features of each signal are extracted and fused to avoid interference from the external working environment of the turbine and to locate the fault. In combination with the above example, when the speed signal and the temperature signal cannot be clearly distinguished, it is necessary to perform feature analysis on the vibration signal. This is because the vibration signal is more sensitive to faults and can be characterized in the early stages of the fault. However, the vibration signal is easily disturbed by the external working environment of the turbine, and the features are not obvious. It is necessary to extract the features of each signal through feature extraction methods and fuse them to achieve fault analysis.

[0053] Specifically, before fusing the various signals in the monitoring signal, the following steps are included:

[0054] (1) Acquiring the monitoring signal; the monitoring signal at least includes a speed signal, a temperature signal and a vibration signal.

[0055] The monitoring signals are collected based on the set sensors, and the data is verified to obtain the verified detection signals.

[0056] (2) Performing a de-averaging operation on the vibration signal.

[0057] Based on the content of independently identifying faults of any of the above signals, when there is no abnormality in the speed signal and the temperature signal, the vibration signal is de-averaged. The de-averaging operation is to eliminate the DC component in the vibration signal, making the signal more stable and easier to analyze.

[0058] (3) Performing data balancing on the vibration signal after removing the mean value and the speed signal.

[0059] It should be noted that the purpose of data balancing is to ensure that the data points of different signals are aligned in time so that joint analysis can be performed.

[0060] (4) Angle resampling is performed on the vibration signal and the rotation speed signal after data balancing, and the effective value and kurtosis of the resampled vibration signal are calculated.

[0061] It should be noted that after the vibration signal and the rotation speed signal are aligned in time, the vibration signal and the rotation speed signal are angle resampled, wherein the angle resampling is to eliminate the influence of the rotation speed fluctuation and convert the signal to a stable angle domain for analysis.

[0062] (5) Perform order spectrum analysis based on the resampled signal to calculate the vibration amplitude corresponding to the rotation frequency order, bearing fault order, and fan blade order.

[0063] It should be noted that the non-stationary signal of variable speed vibration is converted into a stationary signal in the angle domain by order tracking, and the order characteristics are extracted in combination with the fault mechanism, which can effectively reduce the interference of complex environments.

[0064] It should be noted that the method for fusing each signal in the monitoring signal is to calculate numerical parameters of at least two dimensions based on each signal, and to fuse each signal based on the fusion of the numerical parameters, including:

[0065] Acquire monitoring signal characteristics; the monitoring signal characteristics include speed signal characteristics, temperature signal characteristics and vibration signal characteristics after order analysis; wherein, the speed signal characteristics include average speed and speed change rate; the temperature signal characteristics include average temperature of fan end bearing, temperature change rate of fan end bearing, average temperature of turbine end bearing, temperature change rate of turbine end bearing; the vibration signal characteristics after order analysis include effective value and kurtosis of vibration signal after order tracking, rotation frequency order, bearing fault order and vibration amplitude corresponding to fan blade order; calculate numerical parameters of at least two dimensions based on the monitoring signal characteristics, and fuse each signal based on the fusion of the numerical parameters.

[0066] Specifically, the signal features of different dimensions are normalized to the same scale (for example, all features are standardized to a distribution with a mean of 0 and a standard deviation of 1, or normalized to a range of 0 to 1), and then different weights are assigned to different signal features, and each feature is linearly combined according to these weights. For example, the importance weight of each signal feature can be determined based on historical fault data or expert knowledge, and then the weighted sum of each feature is calculated. Finally, different features can be combined into a high-dimensional feature space by constructing a multi-dimensional feature vector. Then, machine learning algorithms are applied in this high-dimensional space for fault detection and classification. Optionally, before fusion, feature selection techniques (such as principal component analysis PCA, linear discriminant analysis LDA, etc.) can be used to reduce the dimensionality of signal features, convert multiple high-dimensional features into a few comprehensive features, and then perform feature fusion based on the comprehensive features.

[0067] Among them, the detailed process of calculating the numerical parameters of at least two dimensions based on the monitoring signal characteristics can be found in the introduction of related technologies and will not be repeated here.

[0068] In particular, under certain conditions (when the characteristics are very obvious), the fault state of the turbo cooler can be identified directly based on the numerical parameters of one dimension. Under such conditions, the fault state of the turbo cooler can be identified more quickly.

[0069] Based on the above content, after obtaining the fused signal, the fused signal is input into the neural network model to output the fault status of the turbine cooler. The process specifically includes:

[0070] A neural network model is established based on the acquired normal data of the turbine cooler and data of different fault degrees of rotor imbalance, blade wear and bearing wear; the neural network model is optimized using a particle swarm optimization algorithm; the fused signal is input into the neural network model, and the fault status of the turbine cooler is output.

[0071] It should be noted that the use of the particle swarm optimization algorithm to optimize the neural network model can be used to search for the optimal initial weights and thresholds of the neural network through the particle swarm optimization algorithm, ensure the convergence and accuracy of the neural network, and greatly improve the accuracy of the diagnosis results of the neural network.

[0072] It should also be noted that the neural network model can be a two-layer neural network model; wherein the first layer of the two-layer neural network model uses a deep neural network to achieve qualitative diagnosis of fault types; and the second layer of the two-layer neural network model uses a preset number of deep neural networks to achieve quantitative diagnosis of different fault degrees under each fault type.

[0073] Figure 2 This is a schematic diagram of the structure of the neural network model shown in Example 1 of the present application. This figure is an exemplary figure. Please refer to Figure 2 ,exist Figure 2 In the example shown, the neural network is a BP neural network. Specifically, the first layer of the network uses a deep BP network to achieve qualitative diagnosis of fault types, and the second layer of the network uses multiple deep BP networks to achieve quantitative diagnosis of different fault severity levels under each fault type. Based on this model, the fused signal is used as input to output the current health status of the turbine cooler (normal, rotor imbalance or blade wear, bearing wear and its severity).

[0074] In this way, the optimization algorithm is used to improve the training accuracy of the BP neural network, and a double-layer network is used, which can not only identify the fault type, but also quantitatively describe the fault degree, which can effectively guide the subsequent maintenance guarantee strategy and achieve the optimal arrangement of guarantee timing and resources.

[0075] In addition, it should be noted that for other turbine faults with unclear mechanisms, it is necessary to automatically extract hidden features through a deep network, and train the CNN network by adding the speed signal, the original vibration spectrum, and the temperature signal to obtain a fault diagnosis model. Among them, by inputting the speed signal, the original vibration spectrum, and the temperature signal into the CNN network together, the fusion of multi-source information can be achieved. The CNN network can comprehensively extract related features from these multi-dimensional data to improve the accuracy of fault diagnosis.

[0076] It should also be noted that after the fused signal is input into the neural network model and the fault status of the turbine cooler is output, a maintenance plan is formulated according to the fault status, specifically including:

[0077] Obtain the fault type of the turbine cooler; the fault type includes operating parameter fault and mechanical component fault; wherein the operating parameter fault includes at least jamming, overspeed, and overtemperature; the mechanical component fault includes at least rotor imbalance, blade wear, and bearing wear; when the fault is an operating parameter fault, stop the machine for inspection; when the fault is a mechanical component fault, implement a corresponding maintenance guarantee plan based on the severity of the fault.

[0078] The online diagnosis method provided in this embodiment can directly determine the sensor type and layout position by determining the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, thereby avoiding the structural limitations, excessive economic costs and information redundancy caused by installing too many sensors, and the problems of insufficient health status recognition ability, inability to provide early warning of faults and affecting system safety caused by installing too few sensors. In addition, the monitoring position is determined according to the fault mechanism, and monitoring at an appropriate position can also improve the accuracy of fault diagnosis. In addition, the priority is given to independently identifying whether the turbine cooler is faulty based on any one of the monitoring signals, which utilizes the signal pairs in the monitoring signal. The fault sensitivity of the signal with the largest amplitude change under the fault state is analyzed. The signal with the largest amplitude change under the fault state has the highest fault sensitivity. The recognition order of each signal can be obtained from high to low sensitivity. The fault can be identified based on this order, which can realize rapid fault identification and improve the efficiency of fault identification. In this way, the monitoring signal is collected in real time and independently identified whether a fault occurs. The problem can be discovered in time at the early stage of the fault, thereby reducing the potential risk of damage and maintenance costs. When any signal has not yet identified the fault, the characteristics of each signal can be extracted and fused to integrate the information of each signal, avoid interference from the external working environment of the turbine, and increase the comprehensiveness and accuracy of the diagnosis. It should also be noted that through this dual mechanism of independent signal identification and fusion signal, false alarms or missed alarms caused by single signal abnormalities can be reduced, and the overall reliability of the system can be improved. Finally, the neural network model is used to predict the fault state of the fused signal, which can better handle complex signal patterns and nonlinear relationships and provide more accurate fault diagnosis results. This method has strong adaptability and can adjust the monitoring signal and processing method according to different turbine coolers and fault characteristics to ensure efficient fault monitoring and diagnosis capabilities under different operating conditions.

[0079] Corresponding to the aforementioned embodiment of a turbine cooler online diagnosis method, the present application also provides an embodiment of a turbine cooler online diagnosis device.

[0080] Figure 3This is a schematic diagram of the structure of the second embodiment of the turbine cooler online diagnostic device provided by this application. Figure 3 , the device provided in this embodiment includes a determination module 310, a collection module 320, an identification module 330 and a processing module 340;

[0081] The determination module 310 is used to determine the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is a signal with the largest amplitude change under the fault state;

[0082] The acquisition module 320 is used to acquire the monitoring signal in real time;

[0083] The identification module 330 is used to independently identify whether the turbine cooler is faulty based on any one of the monitoring signals;

[0084] Wherein, the recognition order of each signal is determined according to the time sequence of the associated faults of each signal in the monitoring signal;

[0085] The processing module 340 is used to fuse the various signals in the monitoring signal when the recognition result is no fault, input the fused signal into the neural network model, and output the fault state of the turbine cooler;

[0086] Wherein, numerical parameters of at least two dimensions are calculated based on each signal, and each signal is fused based on the fusion of the numerical parameters.

[0087] 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.

[0088] 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.

[0089] 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.

[0090] 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 method for online diagnosis of a turbine cooler, characterized in that: The method comprises: Determine the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is a signal with the largest amplitude change under the fault state, and the amplitude change refers to the change significance rate; Collecting the monitoring signal in real time; independently and individually identifying whether the turbine cooler is faulty based on any one of the monitoring signals; Wherein, the recognition order of each signal is determined according to the time sequence of the associated faults of each signal in the monitoring signal; If the identification result is that there is no fault, the various signals in the monitoring signal are fused, the fused signal is input into the neural network model, and the fault state of the turbine cooler is output; wherein, numerical parameters of at least two dimensions are calculated based on each signal, and each signal is fused based on the fusion of the numerical parameters; The method of independently identifying whether the turbine cooler is faulty based on any one of the monitoring signals comprises: Determine a target monitoring signal corresponding to each fault according to the historical fault information of the turbine cooler, wherein the target monitoring signal is a signal with the largest amplitude change under the fault condition; Establishing a corresponding relationship based on the target monitoring signal and corresponding information of the fault; Determine the to-be-confirmed fault corresponding to each signal in the monitoring signal based on the corresponding relationship; Determining a first recognition order of each signal in the monitoring signal according to the occurrence time of the fault to be confirmed; Adjusting the first recognition order according to the amplitude change rate of each signal to obtain a second recognition order; Each signal in the monitoring signal is identified in sequence according to a second identification order, and only one signal is identified each time to determine whether the turbine cooler is faulty.

2. The method according to claim 1, characterized in that After the monitoring signal is collected in real time, the method includes: Obtaining historical operation data, preset thresholds and time information of the monitoring signal; Whether the collected monitoring signal is abnormal is determined based on the historical operation data, the preset threshold and the time information.

3. The method according to claim 1, characterized in that Before fusing each signal in the monitoring signal, the method includes: Acquire the monitoring signal; the monitoring signal at least includes a speed signal, a temperature signal and a vibration signal; Performing a de-averaging operation on the vibration signal; Performing data balancing on the vibration signal after removing the mean value and the rotation speed signal; Performing angle resampling on the vibration signal and the speed signal after data balancing, and calculating the effective value and kurtosis of the resampled vibration signal; The order spectrum analysis is performed based on the resampled signal to calculate the vibration amplitude corresponding to the rotation frequency order, bearing fault order, and fan blade order.

4. The method according to claim 1, characterized in that: The calculating of at least two-dimensional numerical parameters based on the respective signals, and fusing the respective signals based on the fusion of the numerical parameters, comprises: Acquire monitoring signal characteristics; the monitoring signal characteristics include speed signal characteristics, temperature signal characteristics and vibration signal characteristics after order analysis; wherein the speed signal characteristics include average speed and speed change rate; the temperature signal characteristics include average fan end bearing temperature, fan end bearing temperature change rate, turbine end bearing average temperature, turbine end bearing temperature change rate; the vibration signal characteristics after order analysis include effective value and kurtosis of vibration signal after order tracking, rotation frequency order, bearing fault order and vibration amplitude corresponding to fan blade order; Numerical parameters of at least two dimensions are calculated based on the monitoring signal characteristics, and the individual signals are fused based on the fusion of the numerical parameters.

5. The method according to claim 1, characterized in that The step of inputting the fused signal into a neural network model and outputting the fault status of the turbine cooler comprises: A neural network model is established based on the acquired normal data of the turbine cooler and the data of different fault degrees of rotor imbalance, blade wear and bearing wear; Using a particle swarm optimization algorithm to optimize the neural network model; The fused signal is input into the neural network model, and the fault status of the turbine cooler is output.

6. The method according to claim 5, characterized in that The neural network model is a two-layer neural network model; wherein the first layer of the two-layer neural network model uses a deep neural network to achieve qualitative diagnosis of fault types; the second layer of the two-layer neural network model uses a preset number of deep neural networks to achieve quantitative diagnosis of different fault degrees under each fault type.

7. The method according to claim 1, characterized in that After the fused signal is input into the neural network model and the fault status of the turbine cooler is output, a maintenance plan is formulated according to the degree of the fault, specifically including: Obtaining a turbine cooler fault type; the fault type includes an operating parameter fault and a mechanical component fault; wherein the operating parameter fault includes at least a jam, overspeed, and overtemperature; the mechanical component fault includes at least a rotor imbalance, blade wear, and bearing wear; When the fault is an operating parameter fault, the machine is shut down for inspection; when the fault is a mechanical component fault, a corresponding maintenance guarantee plan is implemented based on the severity of the fault.

8. An online diagnostic device for a turbine cooler, characterized in that: The device includes a determination module, a collection module, an identification module and a processing module; wherein, The determination module is used to determine the monitoring signal type and monitoring position based on the fault mechanism of the turbine cooler, wherein each signal in the monitoring signal will undergo abnormal changes in each fault of the turbine cooler, and for each fault, at least one signal in the monitoring signal is a signal with the largest amplitude change under the fault state, and the amplitude change refers to the change significance rate; The acquisition module is used to acquire the monitoring signal in real time; The identification module is used to independently identify whether the turbine cooler is faulty based on any one of the monitoring signals; Wherein, the recognition order of each signal is determined according to the time sequence of the associated faults of each signal in the monitoring signal; The processing module is used to fuse the various signals in the monitoring signal when the recognition result is no fault, input the fused signal into the neural network model, and output the fault state of the turbine cooler; wherein, numerical parameters of at least two dimensions are calculated based on each signal, and each signal is fused based on the fusion of the numerical parameters; The method of independently identifying whether the turbine cooler is faulty based on any one of the monitoring signals comprises: Determine a target monitoring signal corresponding to each fault according to the historical fault information of the turbine cooler, wherein the target monitoring signal is a signal with the largest amplitude change under the fault condition; Establishing a corresponding relationship based on the target monitoring signal and corresponding information of the fault; Determine the to-be-confirmed fault corresponding to each signal in the monitoring signal based on the corresponding relationship; Determining a first recognition order of each signal in the monitoring signal according to the occurrence time of the fault to be confirmed; Adjusting the first recognition order according to the amplitude change rate of each signal to obtain a second recognition order; Each signal in the monitoring signal is identified in sequence according to a second identification order, and only one signal is identified each time to determine whether the turbine cooler is faulty.

Images