A fault diagnosis method, system and related device for a rotating type device

Abstract

The application discloses a rotating device fault diagnosis method and system and related devices, and relates to the field of nuclear power plants, and comprises the following steps: acquiring current parameter vectors of various rotating devices; acquiring an offset vector constraint range; if the offset vector of the current parameter vector and a historical parameter vector does not fall into the offset vector constraint range, inputting the current parameter vector into an LSTM model, an SVM model and a random forest model respectively to obtain multiple fault category probabilities; inputting the multiple fault category probabilities into a CNN model for probability summarization to acquire a fault category corresponding to the current parameter vector. Through the current parameter vector, the application jointly diagnoses the states of various rotating devices, the offset vector constraint range quickly identifies the state of the current parameter vector, and the current parameter vector is predicted through three different models, and then probability summarization is performed, so that the overfitting problem existing in single model prediction is prevented, and the fault prediction accuracy is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of nuclear power plants, and in particular to a fault diagnosis method and system for rotating equipment and related devices. BACKGROUND

[0002] In a nuclear power plant, various rotating equipment (such as motors, pumps, etc.) is critical to the normal operation of the nuclear power plant, and the operational reliability of various rotating equipment can directly affect the safety of the nuclear power plant.

[0003] Since the working states of various rotating equipment affect each other, it is necessary to uniformly diagnose the faults of various rotating equipment to ensure the normal operation between various rotating equipment. SUMMARY

[0004] In view of the above problems, the present application provides a fault diagnosis method and system for rotating equipment and related devices to achieve the purpose of fault diagnosis of various rotating equipment. The specific scheme is as follows:

[0005] The first aspect of the present application provides a fault diagnosis method for rotating equipment, which comprises:

[0006] Respectively acquiring the operating parameters of various rotating equipment at the same time, and constructing a current parameter vector according to the operating parameters of various rotating equipment at the same time;

[0007] Obtaining an offset vector constraint range, if the offset vector of the current parameter vector and the historical parameter vector does not fall into the offset vector constraint range, inputting the current parameter vector into a long short-term memory network (LSTM) model, a support vector machine (SVM) model and a random forest model respectively, and the offset vector constraint range is calculated according to a plurality of historical parameter vectors of the current parameter vector;

[0008] Obtaining a plurality of fault class probabilities of the current parameter vector predicted by the LSTM model, a plurality of fault class probabilities of the current parameter vector predicted by the SVM model, and a plurality of fault class probabilities of the current parameter vector predicted by the random forest model;

[0009] Inputting the plurality of fault class probabilities predicted by the LSTM model, the plurality of fault class probabilities predicted by the SVM model, and the plurality of fault class probabilities predicted by the random forest model into a convolutional neural network (CNN) model for probability summarization, and acquiring a fault class corresponding to the current parameter vector output by the CNN model.

[0010] In one possible implementation, the respective acquisition of the operating parameters of various rotating equipment at the same time comprises:

[0011] Obtain the operating parameters of various rotating devices, perform timestamp alignment processing, and obtain the operating parameters of various rotating devices at the same time.

[0012] In one possible implementation, the calculation process for the offset vector constraint range includes:

[0013] Obtain multiple historical parameter vectors at multiple consecutive moments of the current parameter vector, with each moment corresponding to one of the historical parameter vectors;

[0014] For each target historical parameter vector, a state reference matrix is ​​constructed based on multiple other historical parameter vectors. The similarity between the target historical parameter vector and each other historical parameter vector in the state reference matrix is ​​calculated. Based on the similarity, the weight coefficient of the target historical parameter vector with respect to each other historical parameter vector in the state reference matrix is ​​calculated. Based on the weight coefficient and the state reference matrix, the predicted parameter vector of the target historical parameter vector is calculated. The offset vector between the target historical parameter vector and the predicted parameter vector is obtained. The target historical parameter vector is one of the multiple historical parameter vectors, and the multiple other historical parameter vectors are historical parameter vectors other than the target historical parameter vector.

[0015] Obtain the offset vectors of multiple historical parameter vectors, and calculate the average and deviation values ​​of the multiple offset vectors;

[0016] Based on the average and deviation values ​​of multiple offset vectors, the minimum and maximum values ​​of the offset vectors are calculated to obtain the constraint range of the offset vectors.

[0017] In one possible implementation, the training data for the LSTM model, the SVM model, and the random forest model includes multiple historical parameter vectors labeled with fault categories and multiple historical parameter vectors labeled with no anomalies, and the training datasets for the LSTM model, the SVM model, and the random forest model are all different.

[0018] In one possible implementation, the probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the random forest model are input into a convolutional neural network (CNN) model for probability summarization. The fault category corresponding to the current parameter vector output by the CNN model is then obtained, including:

[0019] The multiple fault category probabilities predicted by the LSTM model, the multiple fault category probabilities predicted by the SVM model, and the multiple fault category probabilities predicted by the random forest model are concatenated into a feature vector, and the feature vector is input into the CNN model, so that the CNN model sequentially performs feature extraction, feature fusion, and category probability transformation on the feature vector;

[0020] Obtain the probabilities of multiple fault categories output by the CNN model, and select the fault category corresponding to the highest probability value as the fault category corresponding to the current parameter vector.

[0021] A second aspect of this application provides a fault diagnosis system for rotating equipment, the fault diagnosis system for rotating equipment comprising:

[0022] The acquisition unit is used to acquire the operating parameters of various types of rotating equipment at the same time, and to construct the current parameter vector based on the operating parameters of various types of rotating equipment at the same time.

[0023] The constraint unit is used to obtain the offset vector constraint range. If the offset vector between the current parameter vector and the historical parameter vector does not fall within the offset vector constraint range, the current parameter vector is input into the Long Short-Term Memory (LSTM) network model, the Support Vector Machine (SVM) model, and the Random Forest model, respectively. The offset vector constraint range is calculated based on multiple historical parameter vectors of the current parameter vector.

[0024] The prediction unit is used to obtain the probability of multiple fault categories of the current parameter vector predicted by the LSTM model, the probability of multiple fault categories of the current parameter vector predicted by the SVM model, and the probability of multiple fault categories of the current parameter vector predicted by the random forest model.

[0025] The classification unit is used to input the probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the random forest model into the convolutional neural network (CNN) model for probability summarization, and to obtain the fault category corresponding to the current parameter vector output by the CNN model.

[0026] In one possible implementation, the acquisition unit acquires the operating parameters of various types of rotating devices at the same moment, specifically configured as follows:

[0027] Obtain the operating parameters of various rotating devices, perform timestamp alignment processing, and obtain the operating parameters of various rotating devices at the same time.

[0028] In one possible implementation, the calculation process for the offset vector constraint range within the constraint element is specifically configured as follows:

[0029] Obtain multiple historical parameter vectors for multiple consecutive time points of the current parameter vector, with each time point corresponding to one historical parameter vector; for each target historical parameter vector, construct a state reference matrix based on multiple other historical parameter vectors, calculate the similarity between the target historical parameter vector and each other historical parameter vector in the state reference matrix, calculate the weight coefficient of the target historical parameter vector for each other historical parameter vector in the state reference matrix based on the similarity, calculate the predicted parameter vector of the target historical parameter vector based on the weight coefficient and the state reference matrix, obtain the offset vector between the target historical parameter vector and the predicted parameter vector, where the target historical parameter vector is one of the multiple historical parameter vectors, and the multiple other historical parameter vectors are historical parameter vectors excluding the target historical parameter vector; obtain the offset vectors of the multiple historical parameter vectors, and calculate the average and deviation values ​​of the multiple offset vectors; based on the average and deviation values ​​of the multiple offset vectors, calculate the minimum and maximum values ​​of the offset vectors to obtain the constraint range of the offset vectors.

[0030] A third aspect of this application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0031] The memory is used to store computer programs;

[0032] The processor is used to execute the computer program so that the electronic device can implement the fault diagnosis method for rotating devices described in the first aspect or any implementation thereof.

[0033] The fourth aspect of this application provides a computer program product including computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the fault diagnosis method for rotating devices described in the first aspect or any implementation thereof.

[0034] Based on the above technical solution, this application provides a fault diagnosis method, system, and related apparatus for rotating equipment. The method acquires the operating parameters of various rotating equipment at the same time and constructs a current parameter vector. It calculates the offset vector constraint range using multiple historical parameter vectors and compares the current parameter vector with the offset vectors of multiple historical parameter vectors within this constraint range to determine if the current parameter vector is an abnormal parameter vector. If the offset vector does not fall within the constraint range, the current parameter vector is determined to be an abnormal parameter vector. These abnormal parameter vectors are then input into an LSTM model, a Support Vector Machine (SVM) model, and a Random Forest model respectively for fault category probability prediction. Finally, a CNN model summarizes the probabilities output by the LSTM, SVM, and Random Forest models to obtain the fault category corresponding to the current parameter vector. This method, by constructing a current parameter vector, can jointly diagnose the state of various rotating equipment. Furthermore, the offset vector constraint range calculated from historical parameter vectors can quickly identify the state of the current parameter vector. By using three different models to predict the fault category of the current parameter vector and then summarizing the probabilities, it prevents overfitting issues that can occur with single-model prediction, effectively improving the fault prediction accuracy of the current parameter vector. Attached Figure Description

[0035] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0036] Figure 1 A flowchart illustrating a fault diagnosis method for rotating equipment provided in an embodiment of this application;

[0037] Figure 2 This is a schematic diagram of the structure of a fault diagnosis system for rotating equipment provided in an embodiment of this application;

[0038] Figure 3 This application provides a hardware structure block diagram of an electronic device.

[0039] Figure label:

[0040] 100 - Acquisition unit; 110 - Constraint unit; 120 - Prediction unit; 130 - Classification unit; 301 - Processing device; 302 - ROM; 303 - RAM; 304 - Bus; 305 - I / O interface; 306 - Input device; 307 - Output device; 308 - Storage device; 309 - Communication device. Detailed Implementation

[0041] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0042] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0043] The terms "first," "second," etc., used in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of units is not necessarily limited to those units, but may include other units not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0044] This application provides a fault diagnosis method for rotating equipment. The fault diagnosis method for rotating equipment according to this application will be described in detail below with reference to the accompanying drawings.

[0045] Reference Figure 1 , Figure 1 This is a flowchart illustrating a fault diagnosis method for rotating equipment provided in an embodiment of this application, as shown below. Figure 1 As shown in the embodiment of this application, a data processing method may include steps S10 to S13, which will be described in detail below.

[0046] S10. Obtain the operating parameters of each type of rotating equipment at the same time, and construct the current parameter vector based on the operating parameters of each type of rotating equipment at the same time.

[0047] Rotating equipment in nuclear power plants refers to key equipment with rotating moving parts. This type of equipment achieves core functions such as energy transfer, fluid transport, and mechanical drive through rotational motion, and is crucial for the normal operation and safe stability of the nuclear power plant. Specifically, it can include motors, pumps, etc. The operating parameters of various rotating equipment obtained in this embodiment can include: primary loop pressure, pressurizer water level, condensate tank level, steam generator water level, primary loop coolant flow rate, secondary loop feedwater flow rate, steam generator outlet pipe pressure, hot end temperature, cold end temperature, coolant pressure, and turbine parameters. To avoid vector misalignment due to differences in sampling frequency and transmission delay of sensors or data sources when acquiring the operating parameters of rotating equipment, this embodiment performs timestamp alignment processing after acquiring the operating parameters of various rotating equipment. This adjusts the timestamps of different sensors or data sources to a unified time base, acquiring the operating parameters of various rotating equipment at the same time. This ensures that the constructed current parameter vector corresponds to the operating parameters of various rotating equipment at the same time, facilitating subsequent data processing and analysis. In the process of timestamp alignment, if a timestamp is missing, it can be supplemented by interpolation or padding. If the timestamp intervals are uneven, the timestamps can be divided into different time windows by time window aggregation to even out the time intervals.

[0048] S11. Obtain the offset vector constraint range. If the offset vector between the current parameter vector and the historical parameter vector does not fall within the offset vector constraint range, then input the current parameter vector into the Long Short-Term Memory (LSTM) network model, the Support Vector Machine (SVM) model, and the Random Forest model, respectively. The offset vector constraint range is calculated based on multiple historical parameter vectors of the current parameter vector.

[0049] S12. Obtain the probabilities of multiple fault categories of the current parameter vector predicted by the LSTM model, and the probabilities of multiple fault categories of the current parameter vector predicted by the SVM model and the random forest model.

[0050] The offset vector constraint range refers to the data fluctuation range calculated based on multiple historical parameter vectors of the current parameter vector. These historical parameter vectors can be constructed from historical operating parameters at multiple consecutive time points, with each time point corresponding to one historical parameter vector. This offset vector constraint range represents the fluctuation range of the operating parameters of rotating equipment. Since this fluctuation range is calculated based on the normal historical operating parameters of the rotating equipment, fluctuations in the operating parameter values ​​within this range can be considered normal fluctuations. If the current operating parameter value exceeds this fluctuation range, it can be considered abnormal. Furthermore, the real-time changes at the current time point allow multiple historical time points to also change in real time, thus the offset vector constraint range can be automatically optimized and updated, improving the recognition accuracy of the current parameter vector.

[0051] In this embodiment, the process of obtaining the offset vector constraint range can be specifically described as steps one through four:

[0052] Step 1: Obtain multiple historical parameter vectors for multiple consecutive time points of the current parameter vector, with each time point corresponding to one historical parameter vector;

[0053] Step 2: For each target historical parameter vector, construct a state reference matrix based on multiple other historical parameter vectors, calculate the similarity between the target historical parameter vector and each other historical parameter vector in the state reference matrix, calculate the weight coefficient of the target historical parameter vector with respect to each other historical parameter vector in the state reference matrix based on the similarity, calculate the predicted parameter vector of the target historical parameter vector based on the weight coefficient and the state reference matrix, and obtain the offset vector between the target historical parameter vector and the predicted parameter vector. The target historical parameter vector is one of multiple historical parameter vectors, and the multiple other historical parameter vectors are historical parameter vectors excluding the target historical parameter vector.

[0054] Step 3: Obtain the offset vectors of multiple historical parameter vectors, and calculate the average and deviation values ​​of the multiple offset vectors;

[0055] Step 4: Calculate the minimum and maximum values ​​of the offset vectors based on the average and deviation values ​​of multiple offset vectors to obtain the constraint range of the offset vectors.

[0056] In this embodiment, historical operating parameters from multiple consecutive time points are first acquired, and then standardized and timestamped to obtain historical time-series data (multiple historical parameter vectors, one for each time point). A target historical parameter vector is selected, and a state reference matrix is ​​constructed using the other historical parameter vectors. The target historical parameter vector is one of the multiple historical parameter vectors, and the other historical parameter vectors are those excluding the target historical parameter vector. The subsequent data processing is explained using the target historical parameter vector as an example; the processing of the other historical parameter vectors is the same as that of the target historical parameter vector.

[0057] After selecting the target historical parameter vector, this embodiment uses a sliding window to segment the historical time-series data composed of multiple other historical parameter vectors when constructing the state reference matrix. Multiple historical parameter vectors within the sliding window constitute a single state vector, and clustering algorithms (such as K-Means or DBSCAN) are used to classify similar states. The classified state vectors are then arranged column-wise to obtain the state reference matrix in this embodiment. Alternatively, in another optional embodiment, this embodiment can directly arrange the transpose of each historical parameter vector column-wise according to its timestamp to obtain the state reference matrix in this embodiment.

[0058] State reference matrix The representation of can be as follows:

[0059] ;

[0060] in, It can represent a state reference matrix; It can represent the state reference matrix The total number of columns, then It can represent the state vector in the first column of the state reference matrix. The state reference matrix can be represented by the first... The state vector of the column.

[0061] Calculate the similarity between the target historical parameter vector and the state reference matrix formed by other historical parameter vectors. The similarity can be represented as follows:

[0062] ;

[0063] in, It can represent the similarity between the target historical parameter vector and the state reference matrix formed by other historical parameter vectors; It can represent a target history parameter vector; It can represent a state reference matrix; It can represent a computation function. ,but , , It can represent an exponential function. It can represent the scaling factor.

[0064] When obtaining the similarity between the target historical parameter vector and the state reference matrix formed by other historical parameter vectors. Then, based on similarity The weighting coefficients between the target historical parameter vector and the state reference matrix formed by other historical parameter vectors are analyzed. Weighting coefficient The representation of can be as follows:

[0065] ;

[0066] in, It can represent the weighting coefficients between the target historical parameter vector and the state reference matrix formed by other historical parameter vectors; It can represent transpose; It can represent The weighting coefficients, It can represent The weighting coefficients, The calculation formula can be expressed as: , It can represent a coefficient; It can be represented as the first in the state reference matrix. Column state vector.

[0067] When obtaining the weighting coefficients between the target historical parameter vector and the state reference matrix formed by other historical parameter vectors. Then, the target historical parameter vector can be calculated. Predicted parameter vector The calculation formula can be shown below:

[0068] ;

[0069] in, It can represent the target history parameter vector The predicted parameter vector; It can represent the target history parameter vector Weighting coefficients between the state reference matrix formed by other historical parameter vectors; It can represent a state reference matrix.

[0070] Obtain the target historical parameter vector Predicted parameter vector Then, the target historical parameter vector With the prediction parameter vector After subtraction, the target historical parameter vector is obtained. offset vector .

[0071] After each historical parameter vector undergoes the above calculations, this embodiment can obtain the offset vectors of multiple historical parameter vectors. The average and deviation values ​​of multiple offset vectors are calculated based on the offset vectors of multiple historical parameter vectors. The calculation formula is as follows:

[0072] ;

[0073] ;

[0074] in, It can represent the average of multiple offset vectors; It can represent the total number of offset vectors; It can represent a coefficient; It can represent the first One offset vector; It can represent the deviation value of the offset vector.

[0075] By obtaining the average and deviation values ​​of multiple offset vectors, the minimum and maximum values ​​of the offset vectors can be calculated, thus obtaining the offset vector constraint range and the minimum value of the offset vector in this embodiment. and the maximum value of the offset vector The formula for calculating can be shown below:

[0076] ;

[0077] ;

[0078] Then the offset vector constraint range The representation of can be as follows:

[0079] .

[0080] In this embodiment, the method for calculating the offset vector between the current parameter vector and the historical parameter vector can be as described above. Specifically, this embodiment can construct a state reference matrix based on multiple historical parameter vectors of the current parameter vector, and calculate the similarity between the current parameter vector and the state reference matrix constructed from the historical parameter vectors. Based on the similarity, the weight coefficient between the current parameter vector and the state reference matrix constructed from the historical parameter vectors is further calculated. The product of the weight coefficient and the state reference matrix constructed from the historical parameter vectors is used as the predicted parameter vector of the current parameter vector. The current parameter vector is subtracted from the predicted parameter vector of the current parameter vector to obtain the offset vector between the current parameter vector and the historical parameter vector.

[0081] The predicted parameter vector of the current parameter vector can represent the predicted current operating data based on data fluctuations or trends in historical operating data. The offset vector constraint range represents the fluctuation range of the operating parameters. Therefore, when the offset vector between the current parameter vector and the predicted parameter vector is not within this offset vector constraint range, it indicates that the data fluctuation between the current operating data and the predicted data from historical operating data has exceeded the normal data fluctuation range, and the current operating data is abnormal. In this embodiment, when the current parameter vector is determined to be an abnormal vector through the offset vector constraint range, it can be input into an LSTM model, an SVM model, and a random forest model for fault category prediction to determine the fault category of rotating equipment. This embodiment uses three different models to predict the fault category of the same current parameter vector, which can effectively improve the prediction accuracy and stability of the fault category.

[0082] Among them, the LSTM (Long Short-Term Memory) model is a special type of recurrent neural network model that effectively learns and remembers long-term dependencies in long sequences by introducing a "gating" mechanism to control the flow of information. The SVM (Support Vector Machine) model is a powerful supervised learning algorithm that uses an optimal hyperplane to separate data points of different classes while maximizing the margin (distance from the hyperplane to the nearest data point), effectively handling linearly and non-linearly separable data. The Random Forest model is an ensemble learning algorithm that improves the accuracy and robustness of the model by constructing multiple decision trees and summing their predictions.

[0083] This embodiment selects the model for fault category prediction from multiple candidate models. The candidate models in this embodiment may include: LSTM model, SVM model, random forest model, and KNN (K-Nearest Neighbors) model. The model selection process can be described as follows:

[0084] Acquire training data, which may include multiple historical parameter vectors labeled with fault categories and multiple historical parameter vectors labeled with no anomalies. Construct multiple training datasets (five in this embodiment) containing different sample data based on the training data. For each candidate model, randomly select one training dataset from the five training datasets as the validation set (remove the labels from the training dataset; when training the next candidate model, relabel it to restore it as the training set). The remaining training dataset is used as the training set for the candidate model (therefore, the training datasets for each candidate model are different). Train the candidate model using the training set, calculate the optimal parameters of the model, and then verify the accuracy of the candidate model using the validation set. Thus, obtain the fault category prediction probability of all candidate models for the training data (specifically, the prediction probability of fault type being no anomaly). Calculate the prediction probability output by each candidate model using a correlation coefficient analysis algorithm to obtain the correlation coefficient between each candidate model. Filter the correlation coefficients between candidate models according to a set correlation coefficient range, and finally determine the model used for fault category prediction in this embodiment.

[0085] The specific process of obtaining the training dataset in this embodiment can be described as follows:

[0086] Obtain historical parameter vectors for multiple historical moments. Retrieve the historical operating parameters of each rotating device at multiple historical moments, perform timestamp alignment, and construct and obtain historical parameter vectors for multiple historical moments.

[0087] Calculate the offset vector at the current moment. Select a historical moment as the current moment. Following the method described above for constructing the state reference matrix, construct a state reference matrix from the historical parameter vectors of multiple historical moments prior to the current moment. Calculate the similarity between the current parameter vector and this state reference matrix. Based on the similarity, further calculate the weight coefficient between the current parameter vector and the state reference matrix. Use the product of the weight coefficient and the state reference matrix as the predicted parameter vector for the current parameter vector. Subtract the predicted parameter vector from the current parameter vector to obtain the offset vector between the current parameter vector and the historical parameter vectors.

[0088] The process calculates the offset vectors for multiple historical moments prior to the current moment. Each time, a historical parameter vector is selected as the target historical parameter vector. The remaining historical parameter vectors are used to construct a state reference matrix, as described above. The similarity between the target historical parameter vector and this state reference matrix is ​​calculated. Based on the similarity, a weighting coefficient between the target historical parameter vector and the state reference matrix is ​​further calculated. The product of the weighting coefficient and the state reference matrix is ​​used as the predicted parameter vector for the target historical parameter vector. The target historical parameter vector is subtracted from its predicted parameter vector to obtain the offset vector. Each historical parameter vector can obtain a corresponding offset vector through this process. Therefore, this embodiment can obtain offset vectors for multiple historical moments.

[0089] Calculate the offset constraint range and update it based on the offset vector at the current time step. Select the offset vectors from multiple historical time steps using a sliding window of fixed length, prior to the current time step. The offset vector of consecutive historical moments ( In a series of consecutive historical moments (where the previous moment before the current moment is the end moment), the current sliding window... The offset vector of consecutive historical moments ( Following the method described above for calculating the average offset and deviation, calculate the current sliding window's... The average and deviation values ​​of the offset vectors at consecutive historical moments are used to obtain the offset constraint range corresponding to the current sliding window. It is then determined whether each vector value in the offset vector at the current moment falls within the offset constraint range corresponding to the current sliding window. If none of the vector values ​​are within this range, the current parameter vector is determined to be in an abnormal operating state, and the corresponding fault category is recorded (locating the abnormal vector value in the current offset vector, determining the corresponding operating parameter, and determining the sensor position based on the operating parameter to determine the fault category). If each vector value is within the offset constraint range corresponding to the current sliding window, the current parameter vector is determined to be in a normal operating state, and the current sliding window is moved one position to the right, incrementing the current moment (the [number]th [time]). The current moment (the first moment) is a historical moment, and the next moment after the current moment (the second moment) is a historical moment. The current time is set to the first time in the sliding window. The first time in the sliding window is deleted, and the time in the sliding window after the move is calculated. The offset vector of consecutive historical moments ( The average and deviation values ​​of the sliding window are calculated, and the offset constraint range is updated to obtain the offset constraint range corresponding to the sliding window after the movement. Specifically, within the sliding window after the movement... The average and deviation values ​​of the offset vectors at consecutive historical moments can be directly obtained from the original current moment (the first one). The offset at each time point is used to update the original average and original deviation values. The update formula can be as follows:

[0090] ;

[0091] ;

[0092] in, It can represent the updated average value; It can represent the original average value; It can represent the current time (the first time). The offset vector at each time point; It can represent the offset vector at the first moment within the sliding window before the movement; It can represent the total number of offset vectors within the sliding window; It can represent the average of the updated offsets; It can represent the original deviation value.

[0093] Training data is obtained by labeling parameter vectors based on recorded information, and a dataset is constructed. Multiple parameter vectors under abnormal operating conditions and multiple parameter vectors under normal operating conditions are obtained. Fault categories are labeled for the parameter vectors under abnormal operating conditions, and the fault category for the parameter vectors under normal operating conditions is labeled as "no abnormality," resulting in multiple training data sets. The same number of parameter vectors are taken from the training data according to the fault category to form a new dataset. This process is repeated multiple times to obtain multiple datasets, each with different training data. The fault categories under abnormal operating conditions may include: coolant leakage, main pump bearing wear, SG heat transfer tube rupture (heat transfer tube in the nuclear power plant steam generator), feedwater system failure, pressurizer safety valve malfunction, and condenser vacuum loss, etc.

[0094] This embodiment selects a predictive model from multiple candidate models using the correlation coefficient, which measures the similarity or difference between model outputs. In this embodiment, the correlation coefficient can be the Pearson correlation coefficient, and its calculation formula is as follows:

[0095] ;

[0096] in, It can represent the correlation coefficient; It can represent a coefficient; It can represent the total number of samples in the training set; Model can be represented For the The predicted probability value of each sample; Representation Model The average predicted probability value for the fault type of no anomaly in all samples. Representation Model For the The predicted probability value of each sample. Representation Model The average predicted probability value for the fault type of no anomaly in all samples.

[0097] Therefore, based on the correlation coefficient, this embodiment selects LSTM, SVM, and Random Forest models as the fault category prediction models from among LSTM, SVM, Random Forest, and KNN models.

[0098] In this embodiment, after inputting the current parameter vector into the trained LSTM model, SVM model, and random forest model respectively, the probabilities of multiple fault categories predicted by the LSTM model, SVM model, and random forest model can be obtained respectively. Each model can output multiple predicted fault categories for the current parameter vector. For example, given a current parameter vector, when input into the LSTM model, SVM model, and random forest model respectively, the LSTM model can output that the probability of the current parameter vector corresponding to a fault of type A is 50%, the probability of a fault of type B is 30%, and the probability of a fault of type C is 20%; the SVM model can output that the probability of the current parameter vector corresponding to a fault of type A is 60%, the probability of a fault of type B is 10%, and the probability of a fault of type C is 30%; and the random forest model can output that the probability of the current parameter vector corresponding to a fault of type A is 65%, the probability of a fault of type B is 20%, and the probability of a fault of type C is 15%.

[0099] S13. Input the probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the random forest model into the convolutional neural network (CNN) model to summarize the probabilities and obtain the fault category corresponding to the current parameter vector output by the CNN model.

[0100] Among them, the CNN (Convolutional Neural Network) model is a deep learning model that automatically extracts features from input data through structures such as convolutional layers, pooling layers, and fully connected layers to achieve classification. In this embodiment, after determining multiple prediction models, the probability prediction results of the multiple prediction models for the fault categories of the training data are obtained, and the actual fault categories to which they belong are labeled. These are then used as training data for the CNN model to train the CNN model.

[0101] Specifically, the process of summarizing probabilities in a trained CNN model can be described as follows:

[0102] The probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the Random Forest model are concatenated into a feature vector. This feature vector is then input into the CNN model, which sequentially performs feature extraction, feature fusion, and class probability transformation (using the softmax function). The multiple fault category probabilities output by the CNN model are obtained, and the fault category corresponding to the highest probability value is selected as the fault category corresponding to the current parameter vector.

[0103] This application provides a fault diagnosis method for rotating equipment. The method acquires the operating parameters of various rotating equipment at the same time and constructs a current parameter vector. It calculates the offset vector constraint range using multiple historical parameter vectors and compares the current parameter vector with the offset vectors of multiple historical parameter vectors within this constraint range to determine if the current parameter vector is an abnormal parameter vector. If the offset vector does not fall within the constraint range, the current parameter vector is determined to be an abnormal parameter vector. These parameters are then input into an LSTM model, a Support Vector Machine (SVM) model, and a Random Forest model to predict the probability of the fault category. Finally, a CNN model summarizes the probabilities output by the LSTM, SVM, and Random Forest models to obtain the fault category corresponding to the current parameter vector. This method, by constructing a current parameter vector, can jointly diagnose the state of various rotating equipment. Furthermore, the offset vector constraint range calculated from historical parameter vectors can quickly identify the state of the current parameter vector. By using three different models to predict the fault category of the current parameter vector and then summarizing the probabilities, it avoids the overfitting problem that can occur with a single model, effectively improving the fault prediction accuracy of the current parameter vector.

[0104] Furthermore, when the rotating equipment is a steam turbine unit, the specific process for fault diagnosis can be described as follows:

[0105] Obtain the current parameter vector of the steam turbine unit, containing its operating parameters at multiple operating times. Construct a probability density function using the current parameter vector. Calculate the probability posterior density function based on the constructed probability density function and the prior probability density function. Integrate the mean and deviation values ​​in the probability posterior density function to obtain the marginal posterior density functions corresponding to the mean and deviation values, respectively. Determine all peak values ​​of the marginal posterior density functions and obtain the statistical values ​​of the mean and deviation values ​​based on the peak values.

[0106] Set window size to Get the error value corresponding to the current parameter vector. Statistical value of the average Statistical values ​​of the deviation values Before statistics Error value at each time point ,in It can represent a parameter, which is the error value at the current time. and before Error value at each time point All values ​​are stored in the window. The window contains the statistical values ​​of all error values ​​and the mean. Subtract the values ​​and accumulate the absolute values ​​to obtain a reference coefficient. ,like exist If the range is within the acceptable range, then the turbine unit can be determined to be in normal condition. exist If the range is within the specified range, it can be determined that the turbine unit is currently in an abnormal state, and a level-two warning will be issued. exist If the condition is within the specified range, it can be determined that the turbine unit is currently in a fault state and an early warning will be issued.

[0107] The probability density function can refer to a probability distribution function used to describe a continuous random variable. The probability density function can provide the probability density of the operating parameters of a steam turbine unit around a certain characteristic value, and its specific form can be expressed as follows:

[0108] ;

[0109] in, It can represent the vector value of the current parameter vector; It can represent the vector values ​​of the prediction parameter vector; It can represent the average value corresponding to the error value (the difference between the vector values ​​of the current parameter vector and the vector values ​​of the predicted parameter vector); It can represent the deviation value corresponding to the error value; exp can represent an exponential function; It can represent the number of vector values; It can represent parameters; Hyperparameters can be represented; their calculation formula can be: .

[0110] The prior probability density function quantifies the parameters before considering any observed data, while the posterior probability density function quantifies the parameters after considering the observed data. Specifically, in this embodiment, the process of obtaining the prior probability density function can be as follows:

[0111] Read the historical operating data of the steam turbine unit and construct the historical time-series vector of the steam turbine, which contains data from multiple moments. Its representation can be as follows: Count the total number of vectors in the historical time series vectors. And set a distance threshold Time interval and embedding dimension (When reconstructing the phase space, the minimum dimension required to map one-dimensional time series data to a higher-dimensional space) reconstructs the historical time series vector into a phase space vector, based on... and Analyze all reconstructed phase space vectors , , It can represent the first A historical time-series vector, Parameters can be represented using phase space vectors. Total number of vectors Distance threshold Time interval and state points Calculate the current distance threshold (the number of distinguishable state points in phase space). And the number of state points are respectively Similarity ratio to 1 and Then subtract the values ​​and use the difference as the similarity change value. .

[0112] Based on the above method for calculating similarity change values, different time intervals are determined. The corresponding similarity change values ​​are calculated, and the time interval corresponding to the first minimum value among the similarity change values ​​is selected as the optimal time interval. The difference between the maximum and minimum similarity change values ​​within this optimal time interval is then calculated. The time length when both the difference and the similarity change value are at their minimum is calculated, and the optimal number of state points is analyzed based on this time length. The reconstructed time series vector is determined based on the optimal time interval and the optimal number of state points. The reconstructed time series vector is then transmitted to a recurrent neural network model for analysis, obtaining the predicted time series vector output by the recurrent neural network model after each training iteration.

[0113] Based on the historical and predicted time-series vectors of the turbine unit, an error value sequence containing multiple time points is calculated. The corresponding average and deviation values ​​are analyzed based on this error value sequence to obtain the average and deviation values ​​after each training iteration. The average value is stored in a first subset, and the deviation values ​​are stored in a second subset. The average and deviation values ​​corresponding to the first and second subsets are calculated, and joint parameters (which may include the joint average and joint deviation values) are calculated using these values. A corresponding probability prior density function is then constructed based on the joint parameters.

[0114] The above describes a fault diagnosis method for rotating equipment provided by the embodiments of this application. The following will describe a system that applies the above-described fault diagnosis method for rotating equipment.

[0115] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of a fault diagnosis system for rotating equipment provided in an embodiment of this application. Figure 2 As shown, the fault diagnosis system for this rotating equipment may include:

[0116] The acquisition unit 100 is used to acquire the operating parameters of various types of rotating equipment at the same time, and to construct the current parameter vector based on the operating parameters of various types of rotating equipment at the same time.

[0117] Constraint unit 110 is used to obtain the offset vector constraint range. If the offset vector between the current parameter vector and the historical parameter vector does not fall within the offset vector constraint range, the current parameter vector is input into the Long Short-Term Memory Network (LSTM) model, the Support Vector Machine (SVM) model, and the Random Forest model, respectively. The offset vector constraint range is calculated based on multiple historical parameter vectors of the current parameter vector.

[0118] Prediction unit 120 is used to obtain the probability of multiple fault categories of the current parameter vector predicted by the LSTM model, the probability of multiple fault categories of the current parameter vector predicted by the SVM model, and the probability of multiple fault categories of the current parameter vector predicted by the random forest model.

[0119] The classification unit 130 is used to input the probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the random forest model into the convolutional neural network (CNN) model for probability summarization, and obtain the fault category corresponding to the current parameter vector output by the CNN model.

[0120] In one possible implementation, the acquisition unit 100 acquires the operating parameters of various types of rotating devices at the same time, which can be specifically configured as follows:

[0121] Obtain the operating parameters of various rotating devices, perform timestamp alignment processing, and obtain the operating parameters of various rotating devices at the same time.

[0122] In one possible implementation, the calculation process for the offset vector constraint range in constraint unit 110 can be specifically configured as follows:

[0123] Obtain multiple historical parameter vectors for multiple consecutive time points of the current parameter vector, with each time point corresponding to one historical parameter vector. For each target historical parameter vector, construct a state reference matrix based on multiple other historical parameter vectors. Calculate the similarity between the target historical parameter vector and each other historical parameter vector in the state reference matrix. Calculate the weight coefficient of the target historical parameter vector relative to each other historical parameter vector in the state reference matrix based on the similarity. Calculate the predicted parameter vector of the target historical parameter vector based on the weight coefficients and the state reference matrix. Obtain the offset vector between the target historical parameter vector and the predicted parameter vector. The target historical parameter vector is one of the multiple historical parameter vectors, and the multiple other historical parameter vectors are historical parameter vectors excluding the target historical parameter vector. Obtain the offset vectors of the multiple historical parameter vectors and calculate the average and deviation values ​​of the multiple offset vectors. Calculate the minimum and maximum values ​​of the offset vectors based on the average and deviation values ​​of the multiple offset vectors to obtain the constraint range of the offset vectors.

[0124] In one possible implementation, the training data for the LSTM, SVM, and random forest models includes multiple historical parameter vectors labeled with fault categories and multiple historical parameter vectors labeled with no anomalies, and the training datasets for the LSTM, SVM, and random forest models are different.

[0125] In one possible implementation, the prediction unit 120 can be specifically configured as follows:

[0126] The probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the Random Forest model are concatenated into a feature vector. This feature vector is then input into the CNN model, which performs feature extraction, feature fusion, and category probability transformation on the feature vector in sequence. The multiple fault category probabilities output by the CNN model are obtained, and the fault category corresponding to the highest probability value is selected as the fault category corresponding to the current parameter vector.

[0127] This application also provides an electronic device in its embodiments. (See reference...) Figure 3The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 3 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0128] like Figure 3 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in ROM 302 or a program loaded into RAM 303 from storage device 308. ROM is a read-only memory, and RAM is a random access memory. When the electronic device is powered on, RAM 303 also stores various programs and data required for the operation of the electronic device. The processing unit 301, ROM 302, and RAM 303 are interconnected via bus 304. An I / O interface 305 is also connected to bus 304; the I / O interface is an input / output interface.

[0129] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, memory cards, hard drives, etc.; and communication devices 309. Communication device 309 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0130] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the fault diagnosis methods for rotating devices provided in this application.

[0131] This application also provides a computer-readable storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the fault diagnosis methods for rotating devices provided in this application.

[0132] It should also be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the system embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0133] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0134] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0135] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0136] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0137] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0138] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A fault diagnosis method for rotating equipment, characterized in that, The fault diagnosis method for rotating equipment includes: Obtain the operating parameters of each type of rotating equipment at the same time, and construct the current parameter vector based on the operating parameters of each type of rotating equipment at the same time; Obtain the offset vector constraint range, which is the normal fluctuation range of the operating parameters of rotating equipment. If the offset vector between the current parameter vector and the historical parameter vector falls within the offset vector constraint range, the current parameter vector is a normal parameter vector. If the offset vector between the current parameter vector and the historical parameter vector does not fall within the offset vector constraint range, the current parameter vector is an abnormal parameter vector. The current parameter vector is then input into the Long Short-Term Memory (LSTM) network model, the Support Vector Machine (SVM) model, and the Random Forest model, respectively. The offset vector constraint range is calculated based on multiple historical parameter vectors of the current parameter vector. Obtain the multiple fault category probabilities of the current parameter vector predicted by the LSTM model, the multiple fault category probabilities of the current parameter vector predicted by the SVM model, and the multiple fault category probabilities of the current parameter vector predicted by the random forest model; The probabilities of multiple fault categories predicted by the LSTM model, the SVM model, and the Random Forest model are concatenated into a feature vector. This feature vector is then input into a Convolutional Neural Network (CNN) model. The CNN model performs feature extraction, feature fusion, and category probability transformation on the feature vector to obtain the multiple fault category probabilities output by the CNN model. The fault category corresponding to the highest probability value among the multiple fault category probabilities output by the CNN model is selected as the fault category corresponding to the current parameter vector.

2. The fault diagnosis method for rotating equipment according to claim 1, characterized in that, The process of acquiring the operating parameters of various types of rotating equipment at the same time includes: Obtain the operating parameters of various rotating devices, perform timestamp alignment processing, and obtain the operating parameters of various rotating devices at the same time.

3. The fault diagnosis method for rotating equipment according to claim 1, characterized in that, The calculation process for the offset vector constraint range includes: Obtain multiple historical parameter vectors at multiple consecutive moments of the current parameter vector, with each moment corresponding to one of the historical parameter vectors; For each target historical parameter vector, a state reference matrix is ​​constructed based on multiple other historical parameter vectors. The similarity between the target historical parameter vector and each other historical parameter vector in the state reference matrix is ​​calculated. Based on the similarity, the weight coefficient of the target historical parameter vector with respect to each other historical parameter vector in the state reference matrix is ​​calculated. Based on the weight coefficient and the state reference matrix, the predicted parameter vector of the target historical parameter vector is calculated. The offset vector between the target historical parameter vector and the predicted parameter vector is obtained. The target historical parameter vector is one of the multiple historical parameter vectors, and the multiple other historical parameter vectors are historical parameter vectors other than the target historical parameter vector. Obtain the offset vectors of multiple historical parameter vectors, and calculate the average and deviation values ​​of the multiple offset vectors; Based on the average and deviation values ​​of multiple offset vectors, the minimum and maximum values ​​of the offset vectors are calculated to obtain the constraint range of the offset vectors.

4. The fault diagnosis method for rotating equipment according to claim 1, characterized in that, The training data for the LSTM model, the SVM model, and the random forest model includes multiple historical parameter vectors labeled with fault categories and multiple historical parameter vectors labeled with no anomalies, and the training datasets for the LSTM model, the SVM model, and the random forest model are all different.

5. A fault diagnosis system for rotating equipment, characterized in that, The fault diagnosis system for rotating equipment includes: The acquisition unit is used to acquire the operating parameters of various types of rotating equipment at the same time, and to construct the current parameter vector based on the operating parameters of various types of rotating equipment at the same time. A constraint unit is used to obtain the offset vector constraint range, which is the normal fluctuation range of the operating parameters of rotating equipment. If the offset vector between the current parameter vector and the historical parameter vector falls within the offset vector constraint range, the current parameter vector is a normal parameter vector. If the offset vector between the current parameter vector and the historical parameter vector does not fall within the offset vector constraint range, the current parameter vector is an abnormal parameter vector. The current parameter vector is then input into the Long Short-Term Memory (LSTM) network model, the Support Vector Machine (SVM) model, and the Random Forest model, respectively. The offset vector constraint range is calculated based on multiple historical parameter vectors of the current parameter vector. The prediction unit is used to obtain the probability of multiple fault categories of the current parameter vector predicted by the LSTM model, the probability of multiple fault categories of the current parameter vector predicted by the SVM model, and the probability of multiple fault categories of the current parameter vector predicted by the random forest model. The classification unit is used to concatenate the multiple fault category probabilities predicted by the LSTM model, the multiple fault category probabilities predicted by the SVM model, and the multiple fault category probabilities predicted by the random forest model into a feature vector. The feature vector is then input into a convolutional neural network (CNN) model. The CNN model performs feature extraction, feature fusion, and category probability transformation on the feature vector to obtain the multiple fault category probabilities output by the CNN model. The fault category corresponding to the maximum probability value among the multiple fault category probabilities output by the CNN model is selected as the fault category corresponding to the current parameter vector.

6. The fault diagnosis system for rotating equipment according to claim 5, characterized in that, The acquisition unit retrieves the operating parameters of various types of rotating devices at the same moment, specifically configured as follows: Obtain the operating parameters of various rotating devices, perform timestamp alignment processing, and obtain the operating parameters of various rotating devices at the same time.

7. The fault diagnosis system for rotating equipment according to claim 5, characterized in that, The calculation process for the offset vector constraint range in the constraint element is specifically configured as follows: Obtain multiple historical parameter vectors for multiple consecutive time points of the current parameter vector, with each time point corresponding to one historical parameter vector; for each target historical parameter vector, construct a state reference matrix based on multiple other historical parameter vectors, calculate the similarity between the target historical parameter vector and each other historical parameter vector in the state reference matrix, calculate the weight coefficient of the target historical parameter vector for each other historical parameter vector in the state reference matrix based on the similarity, calculate the predicted parameter vector of the target historical parameter vector based on the weight coefficient and the state reference matrix, obtain the offset vector between the target historical parameter vector and the predicted parameter vector, where the target historical parameter vector is one of the multiple historical parameter vectors, and the multiple other historical parameter vectors are historical parameter vectors excluding the target historical parameter vector; obtain the offset vectors of the multiple historical parameter vectors, and calculate the average and deviation values ​​of the multiple offset vectors; based on the average and deviation values ​​of the multiple offset vectors, calculate the minimum and maximum values ​​of the offset vectors to obtain the constraint range of the offset vectors.

8. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program to enable the electronic device to implement the fault diagnosis method for rotating devices as described in any one of claims 1 to 4.

9. A computer program product, characterized in that, It includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the fault diagnosis method for rotating devices as described in any one of claims 1 to 4.

Images