Nuclear power rotating equipment vibration fault diagnosis method and device, electronic equipment and medium

Abstract

The invention relates to the technical field of nuclear power equipment operation safety and reliability management, in particular to a nuclear power rotating equipment vibration fault diagnosis method and device, electronic equipment and a medium. According to the nuclear power rotating equipment vibration fault diagnosis method provided by the embodiment of the invention, equipment vibration state parameters of the nuclear power rotating equipment need to be collected firstly, and then variational mode decomposition is carried out on the equipment vibration state parameters to obtain a plurality of intrinsic mode components; fault component screening is carried out on each intrinsic mode component, and an effective mode component is determined from the multiple intrinsic mode components; performing time sequence complexity analysis on the effective modal component to obtain effective component characteristic information; and inputting the effective component characteristic information into a pre-trained vibration fault diagnosis model to obtain a vibration fault diagnosis result corresponding to the nuclear power rotating equipment. Therefore, the vibration fault diagnosis precision and recognition efficiency of the nuclear power rotating equipment can be improved.

Description

Technical Field

[0001] This application relates to the field of nuclear power equipment operation safety and reliability management technology, and in particular to a method and device for diagnosing vibration faults in nuclear power rotating equipment, electronic equipment, and medium. Background Technology

[0002] With the continuous development of nuclear power technology and its supporting equipment, the power and parameters of nuclear power units are constantly increasing, and the equipment structure is gradually becoming more complex and difficult to monitor and analyze accurately. The operating status of the rotating equipment in a nuclear power plant has a significant impact on the entire nuclear power unit. During operation, abnormal changes in various parameters of the rotating equipment can cause considerable damage to the nuclear power unit.

[0003] In related technologies, some artificial intelligence models are used to diagnose vibration faults in rotating nuclear power equipment. However, as the size and structure of rotating nuclear power equipment increase, the number of vibration state parameters collected by sensors at various internal measuring points is also rapidly increasing. Consequently, the number of equipment operating state features that can be extracted from these vibration state parameters is also increasing, forming a relatively complex high-dimensional feature space. This high-dimensional feature space contains a lot of redundant information irrelevant to fault diagnosis, reducing the accuracy and efficiency of vibration fault diagnosis for rotating nuclear power equipment. Summary of the Invention

[0004] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method, apparatus, electronic equipment, and medium for diagnosing vibration faults in nuclear power plant rotating equipment, which can improve the accuracy and identification efficiency of vibration fault diagnosis for nuclear power plant rotating equipment.

[0005] The method for diagnosing vibration faults in nuclear power plant rotating equipment according to a first aspect of this application includes:

[0006] Collect vibration status parameters of rotating equipment in nuclear power plants;

[0007] Variational mode decomposition is performed on the vibration state parameters of the equipment to obtain multiple intrinsic mode components; wherein, each intrinsic mode component is used to characterize different frequency components in the vibration state parameters of the equipment.

[0008] For each intrinsic modal component, a fault component screening is performed to determine the effective modal components from the multiple intrinsic modal components;

[0009] Time series complexity analysis is performed on the effective modal components to obtain the characteristic information of the effective components;

[0010] The effective component characteristic information is input into a pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment.

[0011] According to some embodiments of this application, the variational mode decomposition of the vibration state parameters of the equipment to obtain multiple intrinsic mode components includes:

[0012] Hilbert transform is performed on the vibration state parameters of the equipment to obtain information on multiple frequency components.

[0013] Each frequency component information is modulated onto the corresponding modal fundamental frequency band to obtain the corresponding modal frequency information;

[0014] Bandwidth estimation is performed on the modal frequency information of each modal fundamental frequency band to obtain modal bandwidth information;

[0015] Based on the modal bandwidth information, determine the number of mode decompositions corresponding to the modal frequency information;

[0016] The objective function of mode decomposition is determined based on the preset penalty factor and the number of mode decompositions.

[0017] The objective function of the mode decomposition is solved to obtain multiple intrinsic mode components.

[0018] According to some embodiments of this application, the step of performing fault component screening for each intrinsic modal component and determining the effective modal component from the plurality of intrinsic modal components includes:

[0019] Kurtosis is calculated for each intrinsic mode component to obtain the kurtosis statistic corresponding to each intrinsic mode component;

[0020] The intrinsic modal components whose kurtosis statistics meet the preset screening conditions are determined as the effective modal components.

[0021] According to some embodiments of this application, the step of calculating kurtosis for each intrinsic mode component to obtain the kurtosis statistics corresponding to each intrinsic mode component includes:

[0022] Determine the first and second orders;

[0023] The intrinsic mode components are substituted into a preset cumulative distribution function for calculation to obtain the corresponding cumulative distribution parameters;

[0024] The intrinsic mode components, the cumulative distribution parameters, and the first order are substituted into a preset expectation function to calculate the expected value of the first mode component.

[0025] Substituting the expected value of the modal component and the first order into the preset linear moment calculation function, the first-order linear moment is obtained;

[0026] The intrinsic mode components, the cumulative distribution parameters, and the second order are substituted into the expectation function to calculate the expected value of the second mode component.

[0027] Substituting the expected value of the modal components and the second order into the linear moment calculation function, the second-order linear moment is obtained;

[0028] The kurtosis statistic is determined based on the first linear moment and the second linear moment.

[0029] According to some embodiments of this application, the step of performing time series complexity analysis on the effective modal components to obtain effective component characteristic information includes:

[0030] Two temporal subsequences of the reference dimension are determined from the effective modal components;

[0031] A matching calculation is performed on the two modal temporal subsequences of the baseline dimension to obtain the first matching value;

[0032] From the effective modal components, two modal temporal subsequences with a dimension one order higher than the baseline dimension are re-determined;

[0033] A matching calculation is performed on the two modal temporal subsequences with a dimension one order higher than the baseline dimension to obtain a second matching value;

[0034] The effective component characteristic information is obtained by calculating the sample entropy based on the first matching value and the second matching value.

[0035] According to some embodiments of this application, the length of the effective modal component is the total length of the component, and the matching calculation of the two modal temporal subsequences for the reference dimension to obtain the first matching value includes:

[0036] Scalar distance is calculated between the two modal temporal subsequences in the baseline dimension to determine the corresponding first scalar distance upper limit value;

[0037] The matching probability is calculated based on the total length of the component and the upper limit of the first scalar distance between the modal temporal subsequences to obtain the matching probability of multiple first sub-components in the baseline dimension.

[0038] The first matching value is obtained by averaging the matching probabilities of multiple first sub-components.

[0039] According to some embodiments of this application, the length of the effective modal component is the total length of the component, and the matching calculation for two modal temporal subsequences with a dimension one order higher than the reference dimension to obtain a second matching value includes:

[0040] Scalar distance is calculated between two modal temporal subsequences with a dimension one order higher than the baseline dimension to determine the corresponding upper limit of the second scalar distance.

[0041] The matching probability is calculated based on the total length of the component and the upper limit of the second scalar distance between the modal temporal subsequences to obtain the matching probability of multiple second subcomponents with a dimension one order higher than the baseline dimension.

[0042] The second matching value is obtained by averaging the matching probabilities of multiple second sub-components.

[0043] According to some embodiments of this application, before inputting the effective component characteristic information into a pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment, the method further includes pre-training the vibration fault diagnosis model, specifically including:

[0044] Obtain a training dataset and an original vibration fault diagnosis model; wherein the training dataset includes various equipment vibration sample parameters and sample label information corresponding to each equipment vibration sample parameter;

[0045] Variational mode decomposition is performed on the vibration sample parameters of the equipment to obtain multiple intrinsic mode sample components; wherein, each intrinsic mode sample component is used to characterize different frequency components in the vibration sample parameters of the equipment.

[0046] For each intrinsic modal sample component, a fault component screening is performed to determine the effective modal sample components from the multiple intrinsic modal sample components;

[0047] Time series complexity analysis is performed on the effective modal sample components to obtain the characteristic information of the effective sample components;

[0048] Based on the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device, the original vibration fault diagnosis model is trained by parameter optimization to obtain the pre-trained vibration fault diagnosis model.

[0049] According to some embodiments of this application, the step of performing parameter optimization training on the original vibration fault diagnosis model based on the effective sample component characteristic information and the sample label information corresponding to the vibration sample parameters of each of the devices to obtain the pre-trained vibration fault diagnosis model includes:

[0050] Obtain hyperparameters for various types of models;

[0051] Validation sample information is extracted from the characteristic information of multiple valid sample components, and verification label information matching the validation sample information is extracted from the multiple sample label information.

[0052] Based on the verification sample information and the verification label information corresponding to the verification sample information, hyperparameter optimization is performed on the hyperparameters of various types of models to determine the target hyperparameter combination.

[0053] The original vibration fault diagnosis model is configured based on the target hyperparameter combination, and the vibration fault diagnosis model is trained for performance optimization according to the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device, so as to obtain the pre-trained vibration fault diagnosis model.

[0054] According to some embodiments of this application, the step of optimizing hyperparameters of various types of models based on each of the verification sample information and the verification label information corresponding to the verification sample information, and determining the target hyperparameter combination, includes:

[0055] Different hyperparameter combinations are formed based on the various types of hyperparameter combinations described above;

[0056] Based on the verification sample information and the verification label information corresponding to the verification sample information, the performance of different hyperparameter combinations is verified to obtain experimental hyperparameter combinations.

[0057] Based on the aforementioned combination of experimental hyperparameters, determine the verification reference information;

[0058] Based on the verification reference information, update different hyperparameter combinations, return to perform performance verification on different hyperparameter combinations based on the verification sample information and the verification label information, update the experimental hyperparameter combinations, until the preset hyperparameter optimization conditions are met, and determine the experimental hyperparameter combinations as the target hyperparameter combinations.

[0059] According to some embodiments of this application, the step of performing performance verification on different combinations of hyperparameters based on each of the verification sample information and the verification label information corresponding to the verification sample information to obtain experimental hyperparameter combinations includes:

[0060] For each of the hyperparameter combinations, the vibration fault diagnosis model is configured according to the hyperparameter combination to obtain multiple hyperparameter test models;

[0061] The verification sample information is input into each of the hyperparameter test models for fault diagnosis processing to obtain the corresponding sample test information.

[0062] The test information of each sample and the verification label information are compared to obtain the corresponding accuracy verification result;

[0063] In response to the accuracy verification result satisfying the preset optimization verification conditions, the hyperparameter combination corresponding to the accuracy verification result is determined as the experimental hyperparameter combination.

[0064] According to some embodiments of this application, the step of performing performance optimization training on the vibration fault diagnosis model based on the effective sample component characteristic information and the sample label information corresponding to the vibration sample parameters of each of the aforementioned devices to obtain the pre-trained vibration fault diagnosis model includes:

[0065] The effective sample component characteristic information is input into the vibration fault diagnosis model for fault diagnosis processing to obtain the fault diagnosis results of this round.

[0066] Loss data for this round is obtained by calculating the loss based on the sample label information and the fault diagnosis results of this round.

[0067] Based on the loss data returned in this round, the characteristic information of each effective sample component and the sample label information corresponding to the characteristic information of the effective sample component are input into the vibration fault diagnosis model for fault diagnosis processing until the vibration fault diagnosis model meets the preset performance optimization conditions, and the pre-trained vibration fault diagnosis model is obtained.

[0068] A nuclear power plant rotating equipment vibration fault diagnosis device according to a second aspect embodiment of this application includes:

[0069] The parameter acquisition module is used to collect vibration status parameters of the rotating equipment in nuclear power plants.

[0070] The variational mode decomposition module is used to perform variational mode decomposition on the vibration state parameters of the equipment to obtain multiple intrinsic mode components; wherein, each of the intrinsic mode components is used to characterize different frequency components in the vibration state parameters of the equipment.

[0071] The fault component screening module is used to screen for fault components for each of the intrinsic modal components and determine the effective modal components from the multiple intrinsic modal components.

[0072] The modal component analysis module is used to perform time series complexity analysis on the effective modal components to obtain the characteristic information of the effective components.

[0073] The vibration fault diagnosis module is used to input the effective component characteristic information into a pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment.

[0074] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the vibration fault diagnosis method for nuclear power rotating equipment as described in any one of the embodiments of the first aspect of this application.

[0075] Fourthly, embodiments of this application provide a computer-readable storage medium storing a program that is executed by a processor to implement the nuclear power plant rotating equipment vibration fault diagnosis method as described in any one of the embodiments of the first aspect of this application.

[0076] The nuclear power plant rotating equipment vibration fault diagnosis method, apparatus, electronic device, and medium according to the embodiments of this application have at least the following beneficial effects:

[0077] According to the vibration fault diagnosis method for nuclear power plant rotating equipment according to the embodiments of this application, it is necessary to first collect the vibration state parameters of the nuclear power plant rotating equipment, and then perform variational mode decomposition on the vibration state parameters to obtain multiple intrinsic mode components. Each intrinsic mode component is used to characterize different frequency components in the vibration state parameters. Fault component screening is performed on each intrinsic mode component to determine the effective mode components from the multiple intrinsic mode components. Time series complexity analysis is performed on the effective mode components to obtain the characteristic information of the effective components. The characteristic information of the effective components is input into a pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment. In this way, the accuracy and identification efficiency of vibration fault diagnosis for nuclear power plant rotating equipment can be improved.

[0078] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0079] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0080] Figure 1 A schematic flowchart illustrating the vibration fault diagnosis method for nuclear power plant rotating equipment provided in this application embodiment;

[0081] Figure 2 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0082] Figure 3This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0083] Figure 4 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0084] Figure 5 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0085] Figure 6 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0086] Figure 7 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0087] Figure 8 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0088] Figure 9 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0089] Figure 10 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0090] Figure 11 This is another schematic flowchart of the method for diagnosing vibration faults in nuclear power plant rotating equipment provided in the embodiments of this application;

[0091] Figure 12 This is a schematic diagram of the structure of the nuclear power plant rotating equipment vibration fault diagnosis device provided in the embodiments of this application;

[0092] Figure 13 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0093] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0094] In the description of this application, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0095] In the description of this application, it should be understood that the orientation descriptions, such as up, down, left, right, front, and back, are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0096] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0097] In the description of this application, it should be noted that, unless otherwise explicitly defined, terms such as "setting," "installation," and "connection" should be interpreted broadly. Those skilled in the art can reasonably determine the specific meaning of the above terms in this application based on the specific content of the technical solution. Furthermore, the identification of specific steps in the following text does not imply a limitation on the order of steps or execution logic. The execution order and logic between each step should be understood and inferred from the content described in the embodiments.

[0098] With the continuous development of nuclear power technology and its supporting equipment, the power and parameters of nuclear power units are constantly increasing, and the equipment structure is gradually becoming more complex and difficult to monitor and analyze accurately. The operating status of the rotating equipment in a nuclear power plant has a significant impact on the entire nuclear power unit. During operation, abnormal changes in various parameters of the rotating equipment can cause considerable damage to the nuclear power unit.

[0099] In related technologies, some artificial intelligence models are used to diagnose vibration faults in rotating nuclear power equipment. However, as the size and structure of rotating nuclear power equipment increase, the number of vibration state parameters collected by sensors at various internal measuring points is also rapidly increasing. Consequently, the number of equipment operating state features that can be extracted from these vibration state parameters is also increasing, forming a relatively complex high-dimensional feature space. This high-dimensional feature space contains a lot of redundant information irrelevant to fault diagnosis, reducing the accuracy and efficiency of vibration fault diagnosis for rotating nuclear power equipment.

[0100] This application aims to address at least one of the technical problems existing in the prior art. To this end, this application proposes a method, apparatus, electronic equipment, and medium for diagnosing vibration faults in nuclear power plant rotating equipment, which can improve the accuracy and identification efficiency of vibration fault diagnosis for nuclear power plant rotating equipment.

[0101] Reference Figure 1 The nuclear power plant rotating equipment vibration fault diagnosis method according to the embodiments of this application may include, but is not limited to:

[0102] Step S101: Collect vibration status parameters of the nuclear power plant's rotating equipment;

[0103] Step S102: Perform variational mode decomposition on the equipment vibration state parameters to obtain multiple intrinsic mode components; wherein, each intrinsic mode component is used to characterize different frequency components in the equipment vibration state parameters.

[0104] Step S103: For each intrinsic mode component, a fault component screening is performed to determine the effective mode components from multiple intrinsic mode components.

[0105] Step S104: Perform time series complexity analysis on the effective modal components to obtain the characteristic information of the effective components;

[0106] Step S105: Input the effective component characteristic information into the pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment.

[0107] According to the vibration fault diagnosis method for nuclear power plant rotating equipment shown in steps S101 to S105 of the embodiments of this application, it is necessary to first collect the vibration state parameters of the nuclear power plant rotating equipment, and then perform variational mode decomposition on the vibration state parameters to obtain multiple intrinsic mode components. Each intrinsic mode component is used to characterize different frequency components in the vibration state parameters of the equipment. Fault component screening is performed for each intrinsic mode component to determine the effective mode component from the multiple intrinsic mode components. Time series complexity analysis is performed on the effective mode component to obtain the characteristic information of the effective component. The characteristic information of the effective component is input into a pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment.

[0108] It is worth noting that the embodiments proposed in this application can extract key information from the original vibration state parameters of nuclear power plant rotating equipment, gradually eliminate irrelevant redundant information, and effectively solve the technical problems in vibration fault diagnosis of nuclear power plant rotating equipment by utilizing advanced signal processing and machine learning technologies, thereby improving the accuracy and identification efficiency of vibration fault diagnosis for nuclear power plant rotating equipment.

[0109] In some embodiments, step S101 involves collecting vibration state parameters of the nuclear power plant's rotating equipment.

[0110] It should be noted that step S101 involves the process of collecting vibration state parameters of the rotating equipment in nuclear power plants, providing raw data for subsequent fault diagnosis. In nuclear power plants, rotating equipment such as pumps, turbines, and generators are core components, and their operating status directly affects the safety and reliability of the entire nuclear power plant. Therefore, real-time monitoring of the vibration state of these devices is crucial to ensuring their normal operation.

[0111] It should be understood that collecting equipment vibration state parameters is typically accomplished using multiple sensors deployed at critical locations within the equipment. These sensors detect and record vibration data from the nuclear power plant's rotating equipment. This data can include information such as vibration amplitude, frequency, and changes over time. The data collected by the sensors can be multi-dimensional, covering the vibration of the nuclear power plant's rotating equipment at different locations and in different directions. It is important to emphasize that step S101 is the foundation of the entire fault diagnosis process, providing raw data for analyzing and identifying vibration faults in the nuclear power plant's rotating equipment. Through in-depth analysis of this data, abnormal behavior of the nuclear power plant's rotating equipment can be detected in a timely manner, potential faults can be predicted, and preventative maintenance measures can be taken to ensure the safe and stable operation of the nuclear power plant.

[0112] In step S102 of some embodiments, variational mode decomposition is performed on the equipment vibration state parameters to obtain multiple intrinsic mode components; wherein, each intrinsic mode component is used to characterize different frequency components in the equipment vibration state parameters.

[0113] It should be noted that step S102 describes the process of performing variational mode decomposition on the vibration state parameters collected from the rotating nuclear power plant equipment. This step enables the extraction of key frequency components from complex vibration signals, thereby providing important information for identifying and diagnosing potential faults.

[0114] Variational mode decomposition (VMD) is a signal processing technique that decomposes a multimodal signal (i.e., a signal containing multiple frequency components) into a series of single-mode components, or intrinsic mode functions (IMFs), referred to as IMFs in this embodiment. Each IMF contains a single frequency component or a narrow frequency band of the signal, thereby separating the originally mixed frequency components.

[0115] In this embodiment, multiple intrinsic mode components can be obtained by variational mode decomposition of the equipment vibration state parameters, which contain multiple frequency components. Each intrinsic mode component represents a specific vibration mode in the equipment vibration state parameters. These vibration modes may correspond to the vibration during normal operation of the nuclear power plant rotating equipment, or they may indicate certain faults or abnormal conditions. For example, problems such as imbalance, misalignment, or bearing damage in the nuclear power plant rotating equipment will generate specific vibration frequencies, which can be extracted by variational mode decomposition.

[0116] The intrinsic mode components obtained through variational mode decomposition have clear physical meanings and can characterize the deep-seated features of equipment vibration. These features provide a foundation for further fault analysis; for example, by analyzing the amplitude changes, frequency shifts, or phase changes of specific intrinsic mode components, the specific fault type of the equipment can be inferred.

[0117] Furthermore, variational mode decomposition (VMD) has the advantage of not requiring pre-set filters or spectral analysis; instead, it decomposes the signal directly in the time domain. This reduces information loss caused by improper filter design or spectral leakage. This approach not only improves the accuracy of fault diagnosis but also enhances its applicability and flexibility.

[0118] It should be understood that step S102, through variational mode decomposition, decomposes the complex equipment vibration state parameters into multiple intrinsic mode components, each of which characterizes a different frequency component in the equipment vibration state parameters. This process provides a high-quality data foundation for subsequent fault feature extraction and is a key step in achieving accurate fault diagnosis.

[0119] Reference Figure 2 According to some embodiments of this application, step S102 performs variational mode decomposition on the vibration state parameters of the equipment to obtain multiple intrinsic mode components, which may include, but are not limited to:

[0120] Step S201: Perform Hilbert transformation on the equipment vibration state parameters to obtain information on multiple frequency components;

[0121] Step S202: Modulate each frequency component information to the corresponding modal fundamental frequency band to obtain the corresponding modal frequency information;

[0122] Step S203: Perform bandwidth estimation on the modal frequency information of each modal fundamental frequency band to obtain modal bandwidth information;

[0123] Step S204: Determine the number of mode decompositions corresponding to the mode frequency information based on the modal bandwidth information;

[0124] Step S205: Determine the objective function of mode decomposition based on the preset penalty factor and the number of mode decompositions;

[0125] Step S206: Solve the objective function of mode decomposition to obtain multiple intrinsic mode components.

[0126] In some embodiments, step S201 involves performing a Hilbert transform on the equipment vibration state parameters. The Hilbert transform is a mathematical tool that converts equipment vibration state parameters into analytic signals, thereby obtaining information about multiple frequency components. This step is fundamental to variational mode decomposition, enabling the extraction of individual components containing different frequency information from the original complex equipment vibration state parameters.

[0127] In step S202 of some embodiments, the information of each frequency component obtained through the Hilbert transform is modulated onto the corresponding modal fundamental frequency band. This modulation process associates each frequency component with its corresponding modal fundamental frequency band, thereby obtaining the modal frequency information of each mode. This modal frequency information provides the basis for subsequent fault feature extraction.

[0128] In some embodiments, step S203 involves bandwidth estimation for the modal frequency information of each modal fundamental band. Bandwidth estimation is a crucial step in variational mode decomposition, involving calculating the bandwidth of each mode, i.e., the distribution range of the modal frequency information in the frequency domain. The modal bandwidth information generated in this step is essential for determining the number of mode decompositions.

[0129] In some embodiments, step S204 involves determining the number of mode decompositions corresponding to the mode frequency information based on the mode bandwidth information. This step involves determining how many independent mode components need to be decomposed based on the size and distribution of the mode bandwidth. A reasonable number of mode decompositions helps ensure that variational mode decomposition can effectively capture all important frequency components of the signal while avoiding mode aliasing.

[0130] Step S205 in some embodiments involves determining the modal decomposition objective function based on a preset penalty factor and the number of modal decompositions. It should be noted that the penalty factor is a key parameter in variational modal decomposition, controlling the strength of the constraint on modal bandwidth during the decomposition process. The objective function is the function that needs to be minimized or maximized during variational modal decomposition optimization; it typically includes a penalty term for modal bandwidth and possible other constraints. In some more specific embodiments, the preset number of modal decompositions can be represented as K, and the preset penalty factor can be represented as α.

[0131] In some embodiments, step S206 involves solving the mode decomposition objective function to obtain multiple eigenmode components. This solution process can utilize numerical optimization algorithms, such as the alternating direction multiplier method or other iterative methods. By minimizing the objective function, a set of optimal eigenmode components can be obtained. This solution method not only satisfies bandwidth and orthogonality constraints but also maximizes the capture of useful information from the original signal.

[0132] Steps S201 to S206 in some embodiments of this application achieve variational mode decomposition of equipment vibration state parameters through a series of signal processing steps. This process extracts multiple high-quality intrinsic mode components from the equipment vibration state parameters, providing a basis for fault diagnosis.

[0133] In some embodiments, step S103 involves screening for fault components for each intrinsic modal component to determine the effective modal components from multiple intrinsic modal components.

[0134] It should be noted that step S103 aims to screen the intrinsic mode components obtained from variational mode decomposition, with the goal of identifying and determining the effective mode components that can reflect fault information of the nuclear power plant's rotating equipment from among multiple intrinsic mode components. This step is a crucial step in the fault diagnosis process because it relates to whether signal features directly related to equipment faults can be accurately captured.

[0135] Among the multiple intrinsic modal components obtained after variational modal decomposition, not all intrinsic modal components contain information useful for fault diagnosis. Some intrinsic modal components may mainly reflect the normal operating state of the equipment, while others may contain specific vibration modes caused by faults. Therefore, a method is needed to distinguish these intrinsic modal components and identify those effective modal components that are truly helpful for fault diagnosis.

[0136] In this embodiment, fault component screening for intrinsic modal components requires detailed analysis of the intrinsic modal components to identify whether any fault-indicating features exist. For example, statistics such as kurtosis or skewness can be calculated for each intrinsic modal component, which are sensitive to anomalous patterns. If the statistics of a certain intrinsic modal component show obvious anomalies, it may indicate that the component contains fault-related information.

[0137] In addition, fault component screening can further confirm which intrinsic modal components are truly related to faults and which intrinsic modal components reflect normal operating conditions more by comparing the interrelationships between different intrinsic modal components or analyzing their changes over time.

[0138] It should be understood that once the effective modal components are identified, they will be used for subsequent feature extraction and training of the fault diagnosis model. This means that only those effective modal components considered to contain valuable fault information will be retained, while other intrinsic modal components that do not contain effective information will be excluded. This not only improves the accuracy of fault diagnosis but also increases the efficiency of the diagnostic process, as the focus of analysis is concentrated on the signals most likely to contain fault information. This method improves the targeting and accuracy of fault diagnosis and is a key step in achieving efficient and accurate fault diagnosis.

[0139] Reference Figure 3 According to some embodiments of this application, step S103 performs fault component screening for each intrinsic mode component to determine the effective mode component from multiple intrinsic mode components, which may include, but is not limited to:

[0140] Step S301: Perform kurtosis calculation for each intrinsic mode component to obtain the kurtosis statistics corresponding to each intrinsic mode component.

[0141] Step S302: The intrinsic mode components whose kurtosis statistics meet the preset screening conditions are identified as valid mode components.

[0142] In some embodiments, step S301 requires calculating the kurtosis of the intrinsic modal components. Kurtosis is a statistic used to measure the "tail weight" of a probability distribution, i.e., the frequency of extreme values ​​occurring in the distribution. In vibration signal analysis, kurtosis can be used to identify abnormal or fault features in a signal, as these features often cause the tail of the signal distribution to become heavier. By calculating the kurtosis of each intrinsic modal component, a statistic reflecting the distribution characteristics of the intrinsic modal components can be obtained.

[0143] According to some embodiments of this application, step S301 performs kurtosis calculation for each intrinsic mode component to obtain the kurtosis statistics corresponding to each intrinsic mode component, which may include, but is not limited to:

[0144] Determine the first and second orders;

[0145] The intrinsic mode components are substituted into the preset cumulative distribution function for calculation to obtain the corresponding cumulative distribution parameters;

[0146] The intrinsic modal components, cumulative distribution parameters, and first order are substituted into a preset expectation function to calculate the expected value of the first modal component.

[0147] Substituting the expected values ​​of the modal components and the first order into the preset linear moment calculation function, the first-order linear moment is obtained;

[0148] Substituting the intrinsic mode components, cumulative distribution parameters, and second order into the expectation function, the expected value of the second mode component is obtained.

[0149] Substituting the expected values ​​of the modal components and the second order into the linear moment calculation function, we obtain the second-order linear moment;

[0150] Based on the first and second linear moments, the kurtosis statistics are determined.

[0151] In some embodiments of this application, the process of calculating the kurtosis of each intrinsic mode component is described in detail. Kurtosis is a statistic that measures the tail characteristics of a probability distribution of a real-valued random variable, reflecting the "tail weight" of the distribution. In fault diagnosis, kurtosis can be used to identify abnormal or fault features in a signal, as these features often cause the tail of the signal distribution to become heavier.

[0152] In related technologies, statistical measures such as mean and kurtosis have been used to select effective components. However, traditional statistical measures can only eliminate data with large deviations in standard deviation and mean, which has significant limitations in scenarios with few fault samples. Compared with ordinary kurtosis statistics, the embodiments of this application have better stability and can better resist the influence of noise signals. The specific process of the method for calculating the kurtosis of each intrinsic mode component in the embodiments of this application is as follows:

[0153] First, determine the first and second orders, which are fundamental for subsequent calculations of the cumulative distribution parameters and linear moments. For example, the first order is 2 and the second order is 4.

[0154] Next, each intrinsic mode component is substituted into a preset cumulative distribution function for calculation to obtain the corresponding cumulative distribution parameters. These cumulative distribution parameters can be represented as F(x), which describes the cumulative probability of the intrinsic mode component under different values ​​and is necessary for calculating the expected value and linear moments.

[0155] Subsequently, using these cumulative distribution parameters F(x) and the order r, we substitute them into the preset expectation function E(X).r-k；r The expected values ​​of the first and second modal components were calculated respectively. These expected values ​​reflect the average level of the intrinsic modal components at different orders. The expectation function E(X) is... r-k；r The calculation method for ) can be expressed as:

[0156]

[0157] Then, substituting these expected values ​​and their corresponding orders (first and second orders) into the linear moment calculation function yields the first and second linear moments. Linear moments are important statistics describing the distribution characteristics of random variables. The first linear moment is typically related to the mean, while the second linear moment is related to the variance. The linear moment calculation function is λ. r It can be represented as an r-th order linear moment, specifically as:

[0158]

[0159] Finally, based on these calculated linear moments, kurtosis statistics are determined. Calculating kurtosis statistics involves further processing of the first and second linear moments, ultimately yielding a quantized value that reflects the kurtosis characteristics of each intrinsic mode component. A larger value indicates a heavier tail in the signal, potentially containing more anomalies or fault information.

[0160] For example, if the first order is 2 and the second order is 4, then the first linear moment is represented as λ2 and the second linear moment as λ4. Based on this, the kurtosis statistic LK can be expressed as:

[0161]

[0162] Among them, the value of the kurtosis statistic LK reflects the distribution characteristics of the intrinsic mode components. If the signal does not change significantly, the value of λ2 increases, leading to a decrease in kurtosis; meanwhile, λ4 measures the EX... 4；4 -EX 1；4 The difference between extreme values ​​and EX 3；4 -EX 2；4 The difference between the median values ​​will become very large if the signal has large fluctuations. In some cases, when the number of intrinsic mode components in variational mode decomposition is greater than or equal to 3, the intrinsic mode component with the second largest kurtosis statistic LK can be selected as the effective mode component for signal feature extraction.

[0163] Through the above steps, the kurtosis of each intrinsic mode component can be accurately calculated, providing important statistics for fault component screening. This process not only improves the accuracy of fault diagnosis but also enhances the monitoring capability of equipment health, helping to promptly detect and prevent potential serious faults and accidents.

[0164] In some embodiments, step S302 compares the kurtosis statistic with preset screening criteria to determine which intrinsic mode components (IMCs) should be considered valid IMCs. The preset screening criteria can be determined based on historical data, domain knowledge, and previous analysis results. For example, if the kurtosis statistic of an IMC exceeds a certain threshold, it indicates that the IMC contains fault-related signal features and can therefore be identified as a valid IMC. Setting this threshold requires comprehensive consideration of signal characteristics, fault mode diversity, and tolerance for false alarms and false negatives. This kurtosis-based screening method effectively filters out the most likely fault-related components from multiple IMCs. These valid IMCs can then be used for further feature extraction and analysis, such as calculating sample entropy or other complexity metrics, to obtain more detailed fault feature information.

[0165] As illustrated in steps S301 to S302, by calculating kurtosis statistics and comparing them with preset conditions, effective modal components that truly contain fault information are identified from complex vibration signals. This method not only improves the accuracy of fault diagnosis but also enhances the efficiency of the diagnostic process.

[0166] In some embodiments, step S104 involves performing time series complexity analysis on the effective modal components to obtain effective component characteristic information.

[0167] It should be noted that step S104 requires time series complexity analysis of the identified effective modal components. The purpose of this step is to deeply mine and extract the fault characteristic information contained in the effective components, thereby providing richer and more detailed data support for subsequent fault diagnosis.

[0168] It should be noted that time series complexity analysis is an analytical method designed to quantify the complexity and regularity of time series data. In the field of fault diagnosis, this method can identify subtle changes in the operating state of equipment, which may be early signs of failure. By analyzing the time series complexity of effective modal components, the inherent patterns and abnormal modes of equipment vibration can be revealed.

[0169] Various techniques and algorithms can be employed when analyzing the complexity of time series data. One approach is to calculate sample entropy, a statistic that measures the complexity of a time series by measuring the probability of generating new operating conditions within the effective modal components. Effective modal components with higher sample entropy indicate greater complexity, potentially containing more variations and uncertainties, which is an important characteristic in fault diagnosis. Besides sample entropy, other analytical methods, such as Fourier transform, wavelet transform, or machine learning algorithms, can be used to extract features from the effective modal components. For example, the spectral characteristics of the effective modal components can be analyzed to identify periodic components or anomalous frequency peaks; alternatively, machine learning algorithms, such as random forests or support vector machines, can be used to automatically identify key features within the effective modal components.

[0170] In this embodiment, time-series complexity analysis of the effective modal components yields effective component characteristic information, such as periodicity, frequency components, energy distribution, and nonlinear characteristics. This effective component characteristic information is correlated with specific fault modes of the equipment, providing important clues for fault diagnosis. For example, if the effective component characteristic information reflects an abnormal increase in a certain frequency component, it indicates bearing damage, while if the effective component characteristic information reflects a periodic vibration mode, it may indicate misalignment problems in the equipment.

[0171] It should be understood that the time series complexity analysis in step S104 is a process of in-depth analysis of the effective modal components, aiming to extract and identify feature information related to equipment faults. This method can improve the accuracy and reliability of fault diagnosis, providing a scientific basis for predictive maintenance of rotating nuclear power equipment. This step is a crucial link in achieving accurate fault diagnosis; by revealing the complexity and regularity of equipment vibration states, it provides a solid data foundation for subsequent fault diagnosis model training and prediction.

[0172] Reference Figure 4 According to some embodiments of this application, step S104 performs time series complexity analysis on the effective modal components to obtain effective component characteristic information, which may include, but is not limited to:

[0173] Step S401: Determine two temporal subsequences of the reference dimension from the effective modal components;

[0174] Step S402: Perform matching calculations on the two modal time series subsequences of the baseline dimension to obtain the first matching value;

[0175] Step S403: Re-determine two temporal subsequences of the modalities from the effective modal components, each having a dimension one order higher than the baseline dimension;

[0176] Step S404: Perform matching calculations on the two modal time series subsequences with a dimension one order higher than the baseline dimension to obtain the second matching value;

[0177] Step S405: Calculate the sample entropy based on the first matching value and the second matching value to obtain the effective component characteristic information.

[0178] In some embodiments, step S401 involves determining two modal time-series subsequences of a reference dimension from the effective modal components. The selection of the reference dimension can be based on prior knowledge of the characteristics of the equipment vibration signal and an estimation of the fault characteristic frequency range. By selecting an appropriate reference dimension, it can be ensured that the selected modal time-series subsequences contain fault-related sensitive information.

[0179] In some embodiments, step S402 involves performing a matching calculation on the two modal time-series subsequences in the baseline dimension to obtain a first matching value. Matching calculation is a method for measuring the similarity between two subsequences; it determines the degree of similarity between them by comparing corresponding data points in the subsequences. The first matching value reflects the level of similarity between the two subsequences in the baseline dimension.

[0180] Reference Figure 5 According to some embodiments of this application, in step S402, the length of the effective modal component is the total length of the component. Matching calculations are performed on the two modal temporal subsequences of the reference dimension to obtain the first matching value. This may include, but is not limited to:

[0181] Step S501: Calculate the scalar distance between the two modal time series subsequences in the baseline dimension to determine the corresponding upper limit of the first scalar distance.

[0182] Step S502: Based on the total length of the components and the upper limit of the first scalar distance between the corresponding modal temporal subsequences, the matching probability is calculated to obtain the matching probability of multiple first sub-components in the baseline dimension.

[0183] Step S503: Calculate the mean of the matching probabilities of multiple first sub-components to obtain the first matching value.

[0184] In some embodiments, step S501 involves calculating a scalar distance between two modal temporal subsequences in the baseline dimension to determine the corresponding upper limit of the first scalar distance.

[0185] In some more specific embodiments, for an effective modal component of length L, the two vectors V of lengths l and s are calculated. W (l) and V W The maximum distance d(V) between the scalars corresponding to (s) W (l),V W (s) can be expressed as:

[0186] d(V W (l),V W (s)) = max|v(l + i) - v(s + i)|

[0187] Where v() is time - series data with a total component length of L, l = 1, 2, 3…, L; s = 1, 2, 3…, L; l = 1, 2, 3…, W - 1, and W is the reference dimension of the time series. v(l + i) and v(s + i) are two modal time - series subsequences of the reference dimension. The maximum distance d(V W (l),V W (s)) between the scalars corresponding to the two vectors V W (l) and V W (s) is the upper limit value of the first scalar distance.

[0188] In step S502 of some embodiments, based on the total component length L and the upper limit value d(V W (l),V W (s)) of the first scalar distance between the modal time - series subsequences v(l + i) and v(s + i), calculate the matching probability to obtain multiple first sub - component matching probabilities B W,l (p) of the reference dimension W;

[0189] In some more specific embodiments, for the matching calculation of two modal time - series subsequences v(l + i) and v(s + i) of the reference dimension W, the first sub - component matching probability B W (l) and V W (s) at the W point can be calculated by setting a threshold p, which is expressed as: W,l B

[0190] B W,l (p) = A W,l (p)×(L - W) -1

[0191] Where A W,l (p) represents the number of the upper limit value d(V W (l),V W (s)) < p.

[0192] In step S503 of some embodiments, calculate the mean value based on multiple first sub - component matching probabilities to obtain the first matching value.

[0193] In some more specific embodiments, calculate the mean value based on multiple first sub - component matching probabilities B W,l (p) to obtain the first matching value B W (p), which can be expressed as:

[0194]

[0195] In some embodiments, step S403 requires redefining two temporal subsequences of modality from the effective modal components, each with a dimension one order higher than the baseline dimension. This step aims to explore the similarity of the subsequences at the higher dimension to capture complex patterns that may be overlooked at the lower dimension.

[0196] In some embodiments, step S404 involves performing a matching calculation on two modal time-series subsequences with a dimension one order higher than the baseline dimension to obtain a second matching value. The second matching value reflects the similarity level of the two subsequences at the higher dimension and, compared to the first matching value, can provide more information about the complexity of the time series.

[0197] Reference Figure 6 According to some embodiments of this application, the length of the effective modal component is the total length of the component. Step S404 performs matching calculations on two modal temporal subsequences with a dimension one order higher than the reference dimension to obtain a second matching value, which may include, but is not limited to:

[0198] Step S601: Calculate the scalar distance between two modal time series subsequences with a dimension one order higher than the baseline dimension, and determine the corresponding upper limit of the second scalar distance.

[0199] Step S602: Based on the total length of the components and the upper limit of the second scalar distance between the corresponding modal temporal subsequences, the matching probability is calculated to obtain the matching probability of multiple second sub-components with a dimension one order higher than the baseline dimension.

[0200] Step S603: Calculate the mean of the matching probabilities of multiple second sub-components to obtain the second matching value.

[0201] In some embodiments of this application, the time series complexity analysis process for effective modal components involves matching two modal time series subsequences of a higher-order dimension to obtain a second matching value. This requires calculating the scalar distance between these two higher-order modal time series subsequences. The scalar distance is a metric for measuring the similarity between two subsequences, determined by calculating their differences. Based on these differences, an upper limit, the second scalar distance upper limit, can be determined, representing the maximum allowable difference between the two subsequences. Subsequently, the matching probability is calculated using the total length of the component and the corresponding second scalar distance upper limit. This step involves calculating the matching probability for multiple higher-order dimension sub-components to obtain the matching probability for each sub-component. The matching probability reflects the degree to which the two subsequences can match within a given scalar distance upper limit. Finally, the mean is calculated based on these multiple second sub-component matching probabilities. The purpose of the mean calculation is to summarize the multiple matching probability values ​​to obtain a comprehensive matching value, i.e., the second matching value. This second matching value is an average indicator that measures the degree of matching of subsequences in the next higher dimension. It reflects the complexity and regularity of the time series in that dimension.

[0202] In some more specific embodiments, re-determining two modal temporal subsequences with a dimension one order higher than the baseline dimension from the effective modal components can refer to reconstructing the time series modal temporal subsequences v(l+i) and v(s+i) from W-dimensional vectors to W+1-dimensional vectors, and calculating the second matching value B by referring to the calculation process of the first matching value shown above. W+1 (p).

[0203] In some embodiments, step S405 involves calculating sample entropy based on the first and second matching values ​​to obtain effective component characteristic information. Sample entropy is a statistic that measures the complexity of a time series. It assesses the complexity of a time series by calculating the matching probability of subsequences under different dimensions. By comparing matching values ​​under different dimensions, sample entropy can quantify the irregularities and complexity in a time series, thereby providing important feature information for fault diagnosis.

[0204] Based on the first and second matching values, the sample entropy is calculated to obtain the effective component characteristic information SampEn(W,p,L), which can be expressed as:

[0205] SampEn(W,p,L)=-ln[B W+1 (p)×B W (p)]

[0206] The embodiments shown in steps S401 to S405 of this application, through time-series complexity analysis of the effective modal components, can reveal the inherent laws and abnormal modes of equipment vibration. This process not only improves the accuracy of fault diagnosis but also enhances the monitoring capability of equipment health, helping to promptly detect and prevent potential serious faults and accidents.

[0207] In step S105 of some embodiments, the effective component characteristic information is input into a pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment.

[0208] It should be noted that step S105 requires the effective component characteristic information extracted from the effective modal components to be used as input data. This effective component characteristic information is fed into a pre-trained vibration fault diagnosis model. The function of this vibration fault diagnosis model is to use this input information to predict or identify the vibration fault state of the nuclear power plant's rotating equipment.

[0209] It should be noted that vibration fault diagnosis models are built based on machine learning or deep learning techniques. They can learn from large amounts of historical data and identify patterns associated with different fault states. During the training phase, the vibration fault diagnosis model has been exposed to various types of vibration data and their corresponding labels, such as normal operation, imbalance, and bearing damage. In this way, the vibration fault diagnosis model learns how to distinguish different vibration states based on the input feature information.

[0210] In step S105, the vibration fault diagnosis model receives effective component characteristic information extracted from the effective modal components, such as sample entropy, frequency components, and energy distribution. This effective component characteristic information effectively characterizes the vibration state of the equipment and is directly related to the existence of equipment faults. Based on this, the vibration fault diagnosis model combines this input information with the knowledge learned during training and then outputs a diagnostic result, indicating which category the current vibration state of the nuclear power plant vibration equipment is most likely to belong to. It should be understood that this method not only improves the speed and accuracy of fault detection but also enhances the monitoring capability of equipment health, helping to prevent potential serious faults and accidents.

[0211] Reference Figure 7 According to some embodiments of this application, before inputting the effective component characteristic information into the pre-trained vibration fault diagnosis model in step S105 to obtain the vibration fault diagnosis result corresponding to the nuclear power plant rotating equipment, the method further includes pre-training the vibration fault diagnosis model, specifically including:

[0212] Step S701: Obtain the training dataset and the original vibration fault diagnosis model; wherein, the training dataset includes vibration sample parameters of various equipment and sample label information corresponding to each vibration sample parameter of equipment;

[0213] Step S702: Perform variational mode decomposition on the equipment vibration sample parameters to obtain multiple intrinsic mode sample components; wherein, each intrinsic mode sample component is used to characterize different frequency components in the equipment vibration sample parameters.

[0214] Step S703: For each intrinsic modal sample component, a fault component screening is performed to determine the effective modal sample components from multiple intrinsic modal sample components.

[0215] Step S704: Perform time series complexity analysis on the effective modal sample components to obtain the characteristic information of the effective sample components;

[0216] Step S705: Based on the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device, perform parameter optimization training on the original vibration fault diagnosis model to obtain a pre-trained vibration fault diagnosis model.

[0217] In some embodiments of this application, prior to step S105, a critical preprocessing and training process is involved, the purpose of which is to create a vibration fault diagnosis model that can accurately diagnose vibration faults in nuclear power plant rotating equipment.

[0218] In step S701 of some embodiments, it is necessary to obtain a training dataset and an original vibration fault diagnosis model. The training dataset is the basis for model learning and contains various equipment vibration sample parameters and sample label information corresponding to these parameters. This sample label information indicates the equipment state represented by the sample, such as whether a fault has occurred or the type of fault.

[0219] In step S702 of some embodiments, variational mode decomposition is performed on the vibration sample parameters of the equipment to obtain multiple intrinsic mode sample components. This step uses variational mode decomposition to decompose the complex vibration signal into several simple frequency components. Each intrinsic mode sample component represents a specific frequency range in the signal, thereby making the signal analysis more accurate.

[0220] In some embodiments, step S703 involves screening each intrinsic modal sample component for fault components to identify valid modal sample components containing fault information from a plurality of intrinsic modal sample components. This step is necessary because not all intrinsic modal sample components contain information useful for fault diagnosis. By calculating statistics such as kurtosis and setting thresholds, components most likely to be associated with faults can be selected.

[0221] Step S704 in some embodiments involves performing time series complexity analysis on the effective modal sample components to extract characteristic information of the effective sample components. This step may include calculating sample entropy or other complexity metrics to quantify the complexity of the time series and reveal the operating status of the device and potential failure modes.

[0222] In some embodiments, step S705 involves training the original vibration fault diagnosis model by optimizing its parameters based on the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device. This step involves adjusting the model's parameters to minimize the difference between the predicted output and the actual label. In this way, the model can learn how to accurately predict the vibration fault state of the device based on the input feature information. After training, the pre-trained vibration fault diagnosis model can be used in step S105, that is, to perform fault diagnosis on new device vibration state parameters and output accurate diagnostic results.

[0223] Reference Figure 8 According to some embodiments of this application, step S705, based on the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device, performs parameter optimization training on the original vibration fault diagnosis model to obtain a pre-trained vibration fault diagnosis model, which may include, but is not limited to:

[0224] Step S801: Obtain various types of model hyperparameters;

[0225] Step S802: Extract verification sample information from multiple valid sample component characteristic information, and extract the verification sample information that matches the verification label information from multiple sample label information.

[0226] Step S803: Based on the information of each validation sample and the validation label information corresponding to the validation sample information, perform hyperparameter optimization on various types of model hyperparameters to determine the target hyperparameter combination;

[0227] Step S804: Configure the original vibration fault diagnosis model based on the target hyperparameter combination, and perform performance optimization training on the vibration fault diagnosis model according to the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device, to obtain the pre-trained vibration fault diagnosis model.

[0228] In some embodiments, step S801 involves obtaining various types of model hyperparameters. Model hyperparameters are parameters in a vibration fault diagnosis model that are set before the learning process begins; they typically affect the model's learning algorithm and capabilities. For example, this can include the learning rate, tree depth (for decision tree models), regularization coefficients, etc.

[0229] In some embodiments, step S802 requires extracting validation sample information from multiple valid sample component characteristic information and extracting the sample information that matches the validation sample information from multiple sample label information as validation label information. This step is part of the model validation process, where the validation sample information is used to evaluate the model's performance during training to ensure that the model is not overfitting and can generalize to unseen data.

[0230] In some embodiments, step S803 is a hyperparameter optimization process, in which various types of model hyperparameters are optimized based on validation sample information and their corresponding validation label information to determine the optimal combination of hyperparameters. This process may involve grid search, random search, or more advanced optimization algorithms, such as Bayesian optimization, aiming to find a set of hyperparameter settings that produce better performance on validation sample information.

[0231] Reference Figure 9 According to some embodiments of this application, step S803, based on each verification sample information and the verification label information corresponding to the verification sample information, performs hyperparameter optimization on various types of model hyperparameters to determine the target hyperparameter combination, which may include, but is not limited to:

[0232] Step S901: Different hyperparameter combinations are formed based on various types of hyperparameter combinations;

[0233] Step S902: Based on the information of each verification sample and the verification label information corresponding to the verification sample information, the performance of different hyperparameter combinations is verified to obtain the experimental hyperparameter combinations.

[0234] Step S903: Determine the verification reference information based on the experimental hyperparameter combination;

[0235] Step S904: Update different hyperparameter combinations based on the verification reference information, return to perform performance verification on different hyperparameter combinations based on the verification sample information and verification label information, update the experimental hyperparameter combinations, until the preset hyperparameter optimization conditions are met, and determine the experimental hyperparameter combinations as the target hyperparameter combinations.

[0236] In some embodiments, step S901 involves forming different combinations of hyperparameters based on various types of model hyperparameters. These combinations are typically generated by varying the values ​​of the model hyperparameters, which can be based on grid search, random search, or more complex optimization algorithms. Each combination of hyperparameters represents a different configuration of the vibration fault diagnosis model, affecting the model's learning algorithm and predictive capabilities.

[0237] In some embodiments, step S902 involves using validation sample information and their corresponding validation label information to perform performance validation on these different hyperparameter combinations. This step is used to evaluate the performance of each hyperparameter combination on unseen data. It should be noted that performance validation can train the vibration fault diagnosis model on a training set and then evaluate its accuracy, recall, or other relevant metrics on a validation set. Through this process, a series of experimental hyperparameter combinations can be obtained, each with its own performance metrics.

[0238] Reference Figure 10 According to some embodiments of this application, step S902 performs performance verification on different hyperparameter combinations based on each verification sample information and the verification label information corresponding to the verification sample information to obtain experimental hyperparameter combinations, which may include, but are not limited to:

[0239] Step S1001: For each hyperparameter combination, configure the vibration fault diagnosis model according to the hyperparameter combination to obtain multiple hyperparameter test models;

[0240] Step S1002: Input the verification sample information into each hyperparameter test model for fault diagnosis processing to obtain the corresponding sample test information;

[0241] Step S1003: Compare the test information and verification label information of each sample to obtain the corresponding accuracy verification result;

[0242] Step S1004: In response to the accuracy verification result satisfying the preset optimization verification conditions, the hyperparameter combination corresponding to the accuracy verification result is determined as the experimental hyperparameter combination.

[0243] In step S1001 of some embodiments, a vibration fault diagnosis model is configured according to its specific parameter settings for each set of hyperparameters. This step generates multiple hyperparameter test models, each representing a different learning configuration. The hyperparameters of these models may include learning rate, tree depth (for tree-based models), regularization parameters, etc., which together determine the model's learning strategy and capabilities.

[0244] In some embodiments, step S1002 involves inputting the verification sample information into these hyperparameter test models for fault diagnosis processing. Each model analyzes the input verification sample according to its configuration and outputs corresponding sample test information, i.e., the model's fault diagnosis result for the sample.

[0245] In step S1003 of some embodiments, the sample test information of each model is compared with the corresponding verification label information. The purpose of this step is to evaluate the consistency between the diagnostic results of the hyperparameter test model and the actual labels. By comparison, the accuracy verification result of each hyperparameter test model can be obtained, that is, the proportion of faults correctly diagnosed by the hyperparameter test model.

[0246] In step S1004 of some embodiments, if the accuracy verification result of the hyperparameter test model meets preset optimization verification conditions, then the corresponding hyperparameter combination will be determined as the experimental hyperparameter combination. These conditions may include reaching or exceeding a specific accuracy threshold, or other predetermined performance indicators. A hyperparameter combination that meets the conditions indicates that its configured hyperparameter test model has good diagnostic performance on the verification data, and is therefore selected as the experimental hyperparameter combination for further optimization or direct application to the final configuration of the vibration fault diagnosis model.

[0247] Through the embodiments shown in steps S1001 to S1004, performance verification is performed on each hyperparameter test model, enabling the evaluation and selection of the optimal hyperparameter combination. This process ensures that the vibration fault diagnosis model has high accuracy and reliability in practical applications, thereby contributing to improved diagnostic capabilities for equipment vibration faults.

[0248] In some embodiments, step S903 involves determining validation reference information based on the experimental hyperparameter combinations. The validation reference information may include hyperparameter combinations that performed well during validation, or the average performance of all hyperparameter combinations during validation. This validation reference information serves as a benchmark for the optimization process, used to compare the performance with that of new hyperparameter combinations.

[0249] In some embodiments, step S904 updates different hyperparameter combinations based on the verification reference information, then returns to step S902 to perform performance verification again. This iterative process continues, attempting to improve the hyperparameter combination each time until preset hyperparameter optimization conditions are met. These conditions may include performance metrics reaching a specific threshold, the optimization process reaching a predetermined number of iterations, or computational resource limitations. Once these conditions are met, the experimental hyperparameter combination is determined as the target hyperparameter combination, i.e., the optimal hyperparameter settings.

[0250] The embodiments of this application shown via steps S901 to S902 provide a hyperparameter optimization loop that iteratively evaluates and updates hyperparameter combinations to find an optimal configuration. This method can significantly improve the performance of vibration fault diagnosis models because it allows the vibration fault diagnosis model to adjust to optimal learning algorithms and parameter settings during training. In this way, high accuracy and reliability of the vibration fault diagnosis model can be ensured in practical applications.

[0251] In some more specific embodiments, step S803 optimizes the hyperparameters of various types of models based on each verification sample information and the verification label information corresponding to the verification sample information, and determines the target hyperparameter combination, which can be implemented based on the VMD-XGBoost algorithm.

[0252] It should be noted that the VMD-XGBoost algorithm consists of a variational mode decomposition (VMD) component and an XGBoost component. The VMD component's parameters include the number of mode decompositions, K, and the penalty factor α. The XGBoost component's parameters include the step size eta, the random sampling ratio sub_sample, and the random column ratio colsample_bytree. Before optimizing the parameters, the optimization range for each hyperparameter needs to be set. Different settings for these five model hyperparameters will result in significant variations in the classification performance of VMD-XGBoost.

[0253] To improve the optimization accuracy and convergence speed of model hyperparameters, this application introduces the heuristic algorithm Pathfinder Algorithm (PFA) to optimize multiple model hyperparameters.

[0254] In the heuristic pathfinding optimization algorithm, individuals in the population are divided into pathfinders and population members, where the population members are the hyperparameter combinations. A population member is selected as a pathfinder if it is in a region with a high expected value at any given time.

[0255] At the initial moment of the algorithm, the initial positions of all population members are randomly initialized, and the member with the best position is selected as the pathfinder. In subsequent iterations, the member with the best position is updated as the new pathfinder, and the remaining members move towards the pathfinder.

[0256] In each iteration, the Pathfinder will move before other members, and the Pathfinder's position will be updated as follows:

[0257]

[0258] Where K is the current iteration number, K max The maximum number of iterations, This is the Pathfinder's new location. This is the current location of the Pathfinder. Let r1 be the location of the previous pathfinder, r1 be a random variable that is uniformly generated in the range [0,1], A be a random perturbation vector to prevent the algorithm from getting stuck in a local optimum, and u1 be a random vector that is uniformly generated in the range [-1,1].

[0259] After the Pathfinder's location is updated, the remaining swarm members follow the Pathfinder to update their locations:

[0260]

[0261] R1 = α·r2, R2 = β·r3

[0262]

[0263] D ij =||x i -x j ||

[0264] in This represents the current position of the i-th population member. This represents the current position of the j-th population member. This represents the updated position of the i-th population member. R1 and R2 are two sets of random vectors, where r2 and r3 are uniformly generated random vectors within the range [0,1]. α is the mutual influence coefficient among population members, and β is the pathfinder's leadership coefficient for population members, which is randomly generated within the range [1,2] in each iteration. ε is the perturbation vector designed for population member movement, where u2 is a uniformly generated random vector with values ​​within the range [-1,1]. D ij Let be the distance between the i-th population member and the j-th population member.

[0265] In some embodiments, step S804 involves configuring the original vibration fault diagnosis model based on a determined target hyperparameter combination, and performing performance optimization training using the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of all equipment. The goal of this process is to fine-tune the model parameters of the vibration fault diagnosis model so that it achieves optimal performance on the training data while maintaining good predictive ability for new data. After completing these steps, a pre-trained vibration fault diagnosis model is obtained, which can be used for actual fault diagnosis tasks.

[0266] Reference Figure 11 According to some embodiments of this application, step S804 performs performance optimization training on the vibration fault diagnosis model based on the effective sample component characteristic information and sample label information corresponding to the vibration sample parameters of each device, to obtain a pre-trained vibration fault diagnosis model, which may include, but is not limited to:

[0267] Step S1101: Input the characteristic information of the effective sample components into the vibration fault diagnosis model for fault diagnosis processing to obtain the fault diagnosis results of this round.

[0268] Step S1102: Calculate the loss based on the sample label information and the fault diagnosis results of this round to obtain the loss data for this round;

[0269] Step S1103: Based on the loss data of this round, the characteristic information of each effective sample component and the sample label information corresponding to the characteristic information of the effective sample component are returned and input into the vibration fault diagnosis model for fault diagnosis processing until the vibration fault diagnosis model meets the preset performance optimization conditions, and the pre-trained vibration fault diagnosis model is obtained.

[0270] In step S1101 of some embodiments, the training process of the vibration fault diagnosis model begins by inputting the characteristic information of the effective sample components into the vibration fault diagnosis model. This characteristic information of the effective sample components is extracted from the vibration sample parameters of the equipment and can characterize different aspects of the vibration state of the equipment. The vibration fault diagnosis model uses this information to perform fault diagnosis processing and generate the fault diagnosis result for this round, that is, the model's prediction of the fault state for each sample.

[0271] In some embodiments, step S1102 involves calculating the loss based on the sample label information and the fault diagnosis results of the current round. Loss calculation is a core step in supervised learning, measuring the difference between the model's predicted values ​​and the true values. The loss function may be cross-entropy loss, mean squared error, or other loss functions suitable for classification problems. By calculating the loss, the loss data for this round can be obtained, i.e., the performance index of the vibration fault diagnosis model in the current training round.

[0272] In step S1103 of some embodiments, based on the loss data of the current round, the model will continuously iterate and train, inputting the characteristic information of each effective sample component and its corresponding sample label information into the vibration fault diagnosis model for fault diagnosis processing, until the vibration fault diagnosis model meets the preset performance optimization conditions. These conditions may include the loss value decreasing to a sufficiently low threshold, the accuracy reaching a predetermined target, or after a certain number of training rounds. Through this iterative process, the vibration fault diagnosis model gradually learns how to accurately identify vibration faults from the characteristic information of the effective sample components, thus optimizing its diagnostic capabilities.

[0273] Finally, when the vibration fault diagnosis model meets the preset performance optimization conditions, a pre-trained vibration fault diagnosis model is obtained. This pre-trained vibration fault diagnosis model not only performs well on the training data, but also can generalize to new and unseen data, providing support for practical vibration fault diagnosis tasks.

[0274] In summary, the performance optimization training steps described in these embodiments are a machine learning model training process that improves the diagnostic accuracy and generalization ability of the vibration fault diagnosis model through iterative training and optimization of the loss function. This approach not only improves the speed and accuracy of fault detection but also enhances the understanding of equipment operating status, facilitating real-time monitoring and early warning of equipment health conditions.

[0275] The embodiments shown in steps S801 to S804 of this application cover the entire process from obtaining model hyperparameters to training the vibration fault diagnosis model. This process allows for the systematic optimization of the vibration fault diagnosis model, ensuring its high accuracy and reliability in practical applications. This method not only improves the speed and accuracy of fault detection by the vibration fault diagnosis model but also enhances the model's understanding of equipment operating conditions, facilitating real-time monitoring and early warning of equipment health.

[0276] The pre-training steps shown in steps S701 to S705 of this application embodiment are crucial for constructing an efficient and accurate vibration fault diagnosis model. Through meticulous data preprocessing, feature extraction, and model training, the model can be ensured to have good generalization ability and fault identification ability in practical applications, thereby improving the accuracy of condition monitoring and fault diagnosis of nuclear power plant rotating equipment.

[0277] Reference Figure 12 The nuclear power plant rotating equipment vibration fault diagnosis device according to the second aspect of this application may include, but is not limited to:

[0278] The parameter acquisition module 1201 is used to collect vibration status parameters of the nuclear power plant rotating equipment.

[0279] The variational mode decomposition module 1202 is used to perform variational mode decomposition on the vibration state parameters of the equipment to obtain multiple intrinsic mode components; wherein, each intrinsic mode component is used to characterize different frequency components in the vibration state parameters of the equipment.

[0280] The fault component screening module 1203 is used to screen for fault components for each intrinsic mode component and determine the effective mode components from multiple intrinsic mode components.

[0281] The modal component analysis module 1204 is used to perform time series complexity analysis on the effective modal components to obtain the characteristic information of the effective components.

[0282] The vibration fault diagnosis module 1205 is used to input the effective component characteristic information into the pre-trained vibration fault diagnosis model to obtain the vibration fault diagnosis results corresponding to the nuclear power plant rotating equipment.

[0283] It is evident that the contents of the above embodiments of the nuclear power plant rotating equipment vibration fault diagnosis method are all applicable to the embodiments of this nuclear power plant rotating equipment vibration fault diagnosis device. The specific functions implemented by the embodiments of this nuclear power plant rotating equipment vibration fault diagnosis device are the same as those of the above embodiments of the nuclear power plant rotating equipment vibration fault diagnosis method, and the beneficial effects achieved are also the same as those achieved by the above embodiments of the nuclear power plant rotating equipment vibration fault diagnosis method.

[0284] Reference Figure 13 , Figure 13 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes:

[0285] The processor 1301 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application.

[0286] The memory 1302 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 1302 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1302 and is called and executed by the processor 1301 to execute the nuclear power plant rotating equipment vibration fault diagnosis method of the embodiments of this application.

[0287] The input / output interface 1303 is used to implement information input and output;

[0288] The communication interface 1304 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0289] Bus 1305 transmits information between various components of the device (e.g., processor 1301, memory 1302, input / output interface 1303, and communication interface 1304);

[0290] The processor 1301, memory 1302, input / output interface 1303 and communication interface 1304 are connected to each other within the device via bus 1305.

[0291] This application also provides a computer program product, which includes a computer program. A processor of a computer device reads and executes the computer program, causing the computer device to perform the aforementioned method for diagnosing vibration faults in nuclear power plant rotating equipment.

[0292] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in this disclosure and the foregoing drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “including,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatuses.

[0293] It should be understood that in this disclosure, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, which may include, but is not limited to, any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0294] It should be understood that in the description of the embodiments of this application, "multiple" means two or more, "greater than", "less than", "exceeding" etc. are understood to exclude the number itself, and "above", "below", "within" etc. are understood to include the number itself.

[0295] In the several embodiments provided in this disclosure, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between apparatuses or units, and may be electrical, mechanical, or other forms.

[0296] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0297] Furthermore, the functional units in the various embodiments of this disclosure can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0298] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and may include, but is not limited to, several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this disclosure. The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0299] It should also be understood that the various implementation methods provided in this application can be combined arbitrarily to achieve different technical effects.

[0300] The above is a detailed description of the embodiments of this disclosure. However, this disclosure is not limited to the above embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this disclosure. All such equivalent modifications or substitutions are included within the scope defined by the claims of this disclosure.

Claims

1. A method for diagnosing vibration faults of nuclear power rotating equipment, characterized in that: include: Collect vibration state parameters of nuclear power rotating equipment; Performing variational modal decomposition on the vibration state parameters of the device to obtain a plurality of eigenmodal components; wherein each of the eigenmodal components is used to characterize a different frequency component in the vibration state parameters of the device; Performing fault component screening for each of the intrinsic modal components, and determining effective modal components from a plurality of the intrinsic modal components; Performing time series complexity analysis on the effective modal components to obtain effective component characteristic information; The effective component characteristic information is input into a pre-trained vibration fault diagnosis model to obtain a vibration fault diagnosis result corresponding to the nuclear power rotating equipment.

2. The method according to claim 1, characterized in that: The variational modal decomposition of the vibration state parameters of the device is performed to obtain a plurality of eigenmodal components, including: Performing Hilbert transformation on the vibration state parameters of the device to obtain information on multiple frequency components; Modulating each of the frequency component information to a corresponding modal baseband to obtain corresponding modal frequency information; Performing bandwidth estimation on the modal frequency information of each modal baseband to obtain modal bandwidth information; Determining the modal decomposition number corresponding to the modal frequency information according to the modal bandwidth information; Determining a modal decomposition objective function according to a preset penalty factor and the number of modal decompositions; The modal decomposition objective function is solved to obtain a plurality of the eigenmodal components.

3. The method according to claim 1, characterized in that The step of screening the fault components for each of the eigenmodal components and determining the effective modal component from the multiple eigenmodal components includes: Calculating the kurtosis of each eigenmode component to obtain the kurtosis statistic corresponding to each eigenmode component; The eigenmode components whose kurtosis statistics meet the preset screening conditions are determined as the effective mode components.

4. The method according to claim 3, characterized in that The kurtosis calculation is performed on each of the eigenmode components to obtain the kurtosis statistics corresponding to each of the eigenmode components, including: Determine the first and second orders; Substituting the intrinsic mode component into a preset cumulative distribution function for calculation to obtain corresponding cumulative distribution parameters; Substituting the eigenmode component, the cumulative distribution parameter and the first order into a preset expectation function for calculation to obtain an expected value of the first mode component; Substituting the expected value of the modal component and the first order into a preset linear moment calculation function to obtain a first-order linear moment; Substituting the eigenmode component, the cumulative distribution parameter and the second order into the expected function for calculation to obtain an expected value of the second mode component; Substituting the expected value of the modal component and the second order into the linear moment calculation function to obtain a second-order linear moment; The kurtosis statistic is determined based on the first order linear moment and the second order linear moment.

5. The method according to claim 1, characterized in that The time series complexity analysis is performed on the effective modal component to obtain effective component characteristic information, including: Determining two modal time-series subsequences of reference dimensions from the effective modal components; Performing a matching calculation on the two modal time series subsequences of the reference dimension to obtain a first matching value; Re-determining two of the modal time series subsequences having a dimension one order higher than a reference dimension from the effective modal components; Performing a matching calculation on the two modal time series subsequences having a dimension one order higher than the reference dimension to obtain a second matching value; Sample entropy calculation is performed based on the first matching value and the second matching value to obtain the effective component characteristic information.

6. The method according to claim 5, characterized in that The length of the effective modal component is the total length of the component, and the matching calculation is performed on the two modal time series subsequences of the reference dimension to obtain a first matching value, including: Performing a scalar distance calculation between the two modal time series subsequences of the reference dimension to determine a corresponding first scalar distance upper limit value; Calculating the matching probability based on the total length of the component and the first scalar distance upper limit corresponding to the modal time series subsequence, to obtain a plurality of first sub-component matching probabilities of the reference dimension; The first matching value is obtained by performing an average calculation based on a plurality of the first sub-component matching probabilities.

7. The method according to claim 5, characterized in that The length of the effective modal component is the total length of the component, and the matching calculation is performed on the two modal time series subsequences with a dimension one order higher than the reference dimension to obtain a second matching value, including: Performing a scalar distance calculation between two of the modal time series subsequences having a dimension one order higher than the reference dimension, and determining a corresponding second scalar distance upper limit value; Calculating the matching probability based on the total length of the component and the upper limit value of the second scalar distance corresponding to the modal time series subsequence, to obtain a plurality of second sub-component matching probabilities of a dimension one order higher than the reference dimension; The second matching value is obtained by performing an average calculation based on a plurality of matching probabilities of the second sub-components.

8. The method according to claim 1, characterized in that: Before inputting the effective component characteristic information into a pre-trained vibration fault diagnosis model to obtain a vibration fault diagnosis result corresponding to the nuclear power rotating equipment, the vibration fault diagnosis model is pre-trained, specifically including: Acquire a training data set and an original vibration fault diagnosis model; wherein the training data set includes a plurality of equipment vibration sample parameters and sample label information corresponding to each of the equipment vibration sample parameters; Performing variational modal decomposition on the equipment vibration sample parameters to obtain a plurality of intrinsic modal sample components; wherein each of the intrinsic modal sample components is used to characterize different frequency components in the equipment vibration sample parameters; Performing fault component screening on each of the intrinsic modal sample components, and determining a valid modal sample component from a plurality of the intrinsic modal sample components; Performing time series complexity analysis on the effective modal sample components to obtain effective sample component characteristic information; Based on the effective sample component characteristic information and the sample label information corresponding to each of the equipment vibration sample parameters, parameter optimization training is performed on the original vibration fault diagnosis model to obtain the pre-trained vibration fault diagnosis model.

9. The method according to claim 8, characterized in that The method of performing parameter optimization training on the original vibration fault diagnosis model based on the effective sample component characteristic information and the sample label information corresponding to each of the device vibration sample parameters to obtain the pre-trained vibration fault diagnosis model includes: Get various types of model hyperparameters; Extracting verification sample information from the plurality of valid sample component characteristic information, and extracting the information matching the verification sample information from the plurality of sample label information as verification label information; Based on each of the verification sample information and the verification label information corresponding to the verification sample information, performing hyperparameter optimization on multiple types of model hyperparameters to determine a target hyperparameter combination; The original vibration fault diagnosis model is configured based on the target hyperparameter combination, and the vibration fault diagnosis model is trained for performance optimization according to the effective sample component characteristic information and the sample label information corresponding to each of the device vibration sample parameters to obtain the pre-trained vibration fault diagnosis model.

10. The method according to claim 9, characterized in that The step of performing hyperparameter optimization on multiple types of model hyperparameters based on the verification sample information and the verification label information corresponding to the verification sample information to determine a target hyperparameter combination includes: Forming different hyperparameter combinations based on various types of said hyperparameter combinations; Based on each of the verification sample information and the verification label information corresponding to the verification sample information, performance verification is performed on different hyperparameter combinations to obtain an experimental hyperparameter combination; Determining verification reference information according to the experimental hyperparameter combination; Based on the verification reference information, different hyperparameter combinations are updated, and the performance verification of different hyperparameter combinations based on the verification sample information and the verification label information is returned to update the experimental hyperparameter combination until the preset hyperparameter optimization condition is met, and the experimental hyperparameter combination is determined as the target hyperparameter combination.

11. The method according to claim 10, characterized in that The performing of performance verification on different hyperparameter combinations based on the verification sample information and the verification label information corresponding to the verification sample information to obtain a test hyperparameter combination includes: For each of the hyperparameter combinations, the vibration fault diagnosis model is configured according to the hyperparameter combination to obtain multiple hyperparameter test models; Inputting the verification sample information into each of the hyperparameter test models for fault diagnosis processing to obtain corresponding sample test information; Compare each of the sample test information and the verification label information to obtain a corresponding accuracy verification result; In response to the accuracy verification result satisfying a preset optimization verification condition, the hyperparameter combination corresponding to the accuracy verification result is determined as the experimental hyperparameter combination.

12. The method according to claim 9, characterized in that The method of performing performance optimization training on the vibration fault diagnosis model according to the effective sample component characteristic information and the sample label information corresponding to each of the device vibration sample parameters to obtain the pre-trained vibration fault diagnosis model includes: Inputting the effective sample component characteristic information into the vibration fault diagnosis model to perform fault diagnosis processing to obtain the current round of fault diagnosis results; Calculate the loss based on the sample label information and the current round fault diagnosis result to obtain the current round loss data; Based on the current round of loss data, the characteristic information of each effective sample component and the sample label information corresponding to the effective sample component characteristic information are input into the vibration fault diagnosis model for fault diagnosis processing until the vibration fault diagnosis model meets the preset performance optimization conditions, thereby obtaining the pre-trained vibration fault diagnosis model.

13. A vibration fault diagnosis device for nuclear power rotating equipment, characterized in that: include: A parameter acquisition module is used to collect vibration state parameters of nuclear power rotating equipment; A variational modal decomposition module, used for performing variational modal decomposition on the vibration state parameters of the device to obtain a plurality of eigenmodal components; wherein each of the eigenmodal components is used for characterizing different frequency components in the vibration state parameters of the device; A fault component screening module, used for performing fault component screening on each of the intrinsic modal components, and determining effective modal components from a plurality of the intrinsic modal components; A modal component analysis module, used for performing time series complexity analysis on the effective modal components to obtain effective component characteristic information; The vibration fault diagnosis module is used to input the effective component characteristic information into a pre-trained vibration fault diagnosis model to obtain a vibration fault diagnosis result corresponding to the nuclear power rotating equipment.

14. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method for diagnosing vibration faults of nuclear power rotating equipment as described in any one of claims 1 to 12 is implemented.

15. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the nuclear power rotating equipment vibration fault diagnosis method as described in any one of claims 1 to 12.

Images