A method and system for diagnosing a rotor misalignment fault of a hydropower station and an electronic device

Abstract

The application discloses a kind of hydroelectric power station rotor misalignment fault diagnosis method, system and electronic equipment, belong to fault detection technical field, method includes following steps: S1: initial rotor vibration data set is collected, and initial rotor vibration data set is preprocessed, and rotor data sample set is obtained;S2: for rotor data sample set, construct and train rotor misalignment fault diagnosis model;S3: the latest collected rotor vibration data is as the input of rotor misalignment fault diagnosis model, and hydroelectric power station rotor misalignment fault diagnosis is carried out.The application is aimed at the situation of hydroelectric power station rotor misalignment, and misalignment fault diagnosis model is constructed using neural network, whether misalignment fault occurs in the operation process of rotor can be effectively detected, and the recognition accuracy of rotor misalignment state is improved.

Description

Technical Field

[0001] This invention belongs to the field of fault detection technology, specifically relating to a method, system, and electronic equipment for diagnosing rotor misalignment faults in hydropower stations. Background Technology

[0002] Misalignment faults are prone to occur during the operation of rotating machinery due to inaccurate connections between shafts, mechanical installation errors, and deformation under load. Related data shows that approximately 60% of rotor failures are related to this. When this fault occurs during the operation of rotating machinery, it often causes many highly damaging effects on the equipment, such as oil film instability, premature bearing failure, and coupling deflection. Therefore, this application proposes a method for diagnosing rotor misalignment faults in hydropower stations. Summary of the Invention

[0003] To address the above problems, this invention proposes a method, system, and electronic equipment for diagnosing rotor misalignment faults in hydropower stations.

[0004] The technical solution of the present invention is: a method for diagnosing rotor misalignment faults in hydropower stations, comprising the following steps:

[0005] S1: Collect the initial rotor vibration dataset, preprocess the initial rotor vibration dataset to obtain the rotor data sample set;

[0006] S2: Construct and train a rotor misalignment fault diagnosis model for the rotor data sample set;

[0007] S3: Use the latest collected rotor vibration data as input to the rotor misalignment fault diagnosis model to perform rotor misalignment fault diagnosis in hydropower stations.

[0008] Further, step S1 includes the following sub-steps:

[0009] S11: Collect the initial rotor vibration dataset, and perform a first-stage elimination on the initial rotor vibration dataset to obtain the first rotor vibration dataset;

[0010] S12: Generate the rotor vibration matrix based on the first rotor vibration dataset;

[0011] S13: Perform a second round of elimination on the rotor vibration matrix to obtain the second rotor vibration dataset;

[0012] S14: Normalize each vibration data in the second rotor vibration dataset, and use the normalized vibration data as the rotor data sample set.

[0013] Further, in step S11, the specific method for performing one round of elimination is as follows: Arrange the rotor vibration data of the initial rotor vibration dataset from smallest to largest to obtain a rotor vibration data sequence; select three rotor vibration data points and divide the rotor vibration data sequence into four equal parts; determine the abnormal rotor vibration data interval based on the three rotor vibration data points. N 1, N u ], which belongs to the abnormal rotor vibration data range [ N 1, N u The rotor vibration data was removed to obtain the first rotor vibration dataset; among which, the lower limit of the abnormal rotor vibration data range was... N 1 and upper limit value N u The calculation formulas are as follows:

[0014]

[0015]

[0016] in, n 1 indicates the minimum value of the selected rotor vibration data. n 2 indicates the maximum value of the selected rotor vibration data. ω 1 indicates the first weight. ω 2 indicates the second weight.

[0017] Furthermore, in step S12, the specific method for generating the rotor vibration matrix is ​​as follows: based on the various rotor vibration data in the first rotor vibration dataset, randomly generate... m × n rotor vibration matrix X Among them, the rotor vibration matrix X Fill in the empty spaces with zeros. m This indicates the row number of the rotor vibration matrix. n This indicates the number of columns in the rotor vibration matrix.

[0018] Further, step S13 includes the following sub-steps:

[0019] S131: Randomly select a column from the rotor vibration matrix and determine the elimination threshold corresponding to that column;

[0020] S132: Remove rotor vibration data that are less than the removal threshold from the column randomly selected in step S131;

[0021] S133: Repeat step S131 to determine the rejection threshold for each column, and repeat step S132 to remove rotor vibration data in each column that are less than the rejection threshold, and use the remaining rotor vibration data as the second rotor vibration dataset.

[0022] Furthermore, in step S131, the threshold for elimination is... T The calculation formula is:

[0023]

[0024] in, x j Represents the rotor vibration matrix X A column vector randomly selected from the data. α Represents a random number between 0 and 1.

[0025] Furthermore, in step S2, the rotor misalignment fault diagnosis model includes a convolutional layer, a fully connected layer, and an output layer;

[0026] The convolutional layer is used to extract various features from the rotor data sample set and generate feature differences; the fully connected layer is used to obtain feature vectors based on the feature differences; and the output layer is used to adjust the parameters of the rotor misalignment fault diagnosis model based on the feature vectors.

[0027] Furthermore, the convolutional layers generate feature differences. D ( p The formula for calculating ) is:

[0028]

[0029] In the formula, sim (·) represents the similarity function. Y (·) represents the characteristics of the rotor data sample set. Y This indicates the number of features in the rotor data sample set.

[0030] Furthermore, the fully connected layer uses the vector transformation loss function to obtain the feature vector, where the vector transformation loss function... c The expression is:

[0031]

[0032] In the formula, M Indicates the number of feature differences. D ( p ) indicates differences in characteristics. α m The mean of the differences in characteristics. β m This represents hyperparameters.

[0033] Furthermore, the specific method for adjusting the parameters of the rotor misalignment fault diagnosis model in the output layer is as follows: the softmax function of the output layer is used to classify the feature vectors to obtain the classification probability of each class, and the parameters of the rotor misalignment fault diagnosis model are updated layer by layer according to the classification probability of each class.

[0034] The present invention also provides a fault diagnosis system for rotor misalignment in hydropower stations, comprising a data acquisition and processing module, a model training module and a fault diagnosis module connected in sequence;

[0035] The data acquisition and processing module is used to acquire the initial rotor vibration dataset, preprocess the initial rotor vibration dataset to obtain the rotor data sample set; the model training module is used to construct and train the rotor misalignment fault diagnosis model for the rotor data sample set; the fault diagnosis module is used to use the latest acquired rotor vibration data as the input of the rotor misalignment fault diagnosis model to perform rotor misalignment fault diagnosis in hydropower stations.

[0036] Furthermore, the data acquisition and processing module employs several vibration signal acquisition sensors.

[0037] Furthermore, several vibration signal acquisition sensors are fixedly installed on the rotor of the hydropower station to collect vibration data of the rotor and obtain an initial rotor vibration dataset.

[0038] Furthermore, the model training module is equipped with a computing power resource submodule, which is used to classify the feature vector using the softmax function of the output layer, obtain the classification probability of each class, and update the parameters of the rotor misalignment fault diagnosis model layer by layer according to the classification probability of each class.

[0039] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of a method for diagnosing rotor misalignment faults in hydropower stations.

[0040] The beneficial effects of this invention are:

[0041] (1) The present invention preprocesses the collected data to overcome the defect that the rotor misalignment diagnosis results are easily affected by the collected data and are inaccurate, and greatly optimizes the construction process of the fault diagnosis model;

[0042] (2) In view of the situation of rotor misalignment in hydropower stations, the present invention uses a neural network to construct a misalignment fault diagnosis model, which can effectively detect whether the rotor has misalignment fault during operation and improve the accuracy of rotor misalignment state identification. Attached Figure Description

[0043] Figure 1 This is a flowchart of a method for diagnosing rotor misalignment faults in hydropower stations.

[0044] Figure 2 This is a schematic diagram of a rotor misalignment fault diagnosis system for a hydropower station. Detailed Implementation

[0045] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0046] like Figure 1 As shown, the present invention provides a method for diagnosing rotor misalignment faults in hydropower stations, comprising the following steps:

[0047] S1: Collect the initial rotor vibration dataset, preprocess the initial rotor vibration dataset to obtain the rotor data sample set;

[0048] S2: Construct and train a rotor misalignment fault diagnosis model for the rotor data sample set;

[0049] S3: Use the latest collected rotor vibration data as input to the rotor misalignment fault diagnosis model to perform rotor misalignment fault diagnosis in hydropower stations.

[0050] In this embodiment of the invention, step S1 includes the following sub-steps:

[0051] S11: Collect the initial rotor vibration dataset, and perform a first-stage elimination on the initial rotor vibration dataset to obtain the first rotor vibration dataset;

[0052] S12: Generate the rotor vibration matrix based on the first rotor vibration dataset;

[0053] S13: Perform a second round of elimination on the rotor vibration matrix to obtain the second rotor vibration dataset;

[0054] S14: Normalize each vibration data in the second rotor vibration dataset, and use the normalized vibration data as the rotor data sample set.

[0055] In this embodiment of the invention, the specific method for performing a first elimination in step S11 is as follows: Arrange the rotor vibration data of the initial rotor vibration dataset from smallest to largest to obtain a rotor vibration data sequence; select three rotor vibration data points and divide the rotor vibration data sequence into four equal parts; determine the abnormal rotor vibration data interval based on the three rotor vibration data points. N 1, N u ], which belongs to the abnormal rotor vibration data range [ N 1, N u The rotor vibration data was removed to obtain the first rotor vibration dataset; among which, the lower limit of the abnormal rotor vibration data range was... N 1 and upper limit value N u The calculation formulas are as follows:

[0056]

[0057]

[0058] in, n 1 indicates the minimum value of the selected rotor vibration data. n 2 indicates the maximum value of the selected rotor vibration data. ω 1 indicates the first weight. ω 2 indicates the second weight.

[0059] In this embodiment of the invention, the specific method for generating the rotor vibration matrix in step S12 is as follows: randomly generating the rotor vibration matrix based on the rotor vibration data of the first rotor vibration dataset. m × n rotor vibration matrix X Among them, the rotor vibration matrix X Fill in the empty spaces with zeros. m This indicates the row number of the rotor vibration matrix. n This indicates the number of columns in the rotor vibration matrix.

[0060] In this embodiment of the invention, step S13 includes the following sub-steps:

[0061] S131: Randomly select a column from the rotor vibration matrix and determine the elimination threshold corresponding to that column;

[0062] S132: Remove rotor vibration data that are less than the removal threshold from the column randomly selected in step S131;

[0063] S133: Repeat step S131 to determine the rejection threshold for each column, and repeat step S132 to remove rotor vibration data in each column that are less than the rejection threshold, and use the remaining rotor vibration data as the second rotor vibration dataset.

[0064] In this embodiment of the invention, in step S131, the threshold for elimination is... T The calculation formula is:

[0065]

[0066] in, x j Represents the rotor vibration matrix X A column vector randomly selected from the data. α Represents a random number between 0 and 1.

[0067] In this embodiment of the invention, in step S2, the rotor misalignment fault diagnosis model includes a convolutional layer, a fully connected layer, and an output layer;

[0068] The convolutional layer is used to extract various features from the rotor data sample set and generate feature differences; the fully connected layer is used to obtain feature vectors based on the feature differences; and the output layer is used to adjust the parameters of the rotor misalignment fault diagnosis model based on the feature vectors.

[0069] In this embodiment of the invention, the convolutional layer generates feature differences. D ( p The formula for calculating ) is:

[0070]

[0071] In the formula, sim (·) represents the similarity function. Y (·) represents the characteristics of the rotor data sample set. Y This indicates the number of features in the rotor data sample set.

[0072] In this embodiment of the invention, the fully connected layer uses a vector transformation loss function to obtain feature vectors, wherein the vector transformation loss function... c The expression is:

[0073]

[0074] In the formula, M Indicates the number of feature differences. D ( p ) indicates differences in characteristics. α m The mean of the differences in characteristics. β m This represents hyperparameters.

[0075] In this embodiment of the invention, the specific method for adjusting the parameters of the rotor misalignment fault diagnosis model in the output layer is as follows: the softmax function of the output layer is used to classify the feature vectors to obtain the classification probability of each class, and the parameters of the rotor misalignment fault diagnosis model are updated layer by layer according to the classification probability of each class.

[0076] like Figure 2 As shown in the embodiment of the present invention, a fault diagnosis system for rotor misalignment in a hydropower station includes a data acquisition and processing module, a model training module, and a fault diagnosis module connected in sequence.

[0077] The data acquisition and processing module is used to acquire the initial rotor vibration dataset, preprocess the initial rotor vibration dataset to obtain the rotor data sample set; the model training module is used to construct and train the rotor misalignment fault diagnosis model for the rotor data sample set; the fault diagnosis module is used to use the latest acquired rotor vibration data as the input of the rotor misalignment fault diagnosis model to perform rotor misalignment fault diagnosis in hydropower stations.

[0078] The data acquisition and processing module uses several vibration signal acquisition sensors.

[0079] Several vibration signal acquisition sensors are fixedly installed on the rotor of the hydropower station to collect vibration data of the rotor and obtain an initial rotor vibration dataset.

[0080] The model training module is equipped with a computing power resource submodule, which is used to classify the feature vector using the softmax function of the output layer, obtain the classification probability of each class, and update the parameters of the rotor misalignment fault diagnosis model layer by layer according to the classification probability of each class.

[0081] In an embodiment of the present invention, an electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of a method for diagnosing rotor misalignment faults in hydropower stations.

[0082] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for diagnosing rotor misalignment faults in hydropower stations, characterized in that, Includes the following steps: S1: Collect the initial rotor vibration dataset, preprocess the initial rotor vibration dataset to obtain the rotor data sample set; step S1 includes the following sub-steps: S11: Collect the initial rotor vibration dataset, and perform a first-stage elimination on the initial rotor vibration dataset to obtain the first rotor vibration dataset; In step S11, the specific method for performing one elimination is as follows: Arrange the rotor vibration data of the initial rotor vibration dataset from smallest to largest to obtain a rotor vibration data sequence; select three rotor vibration data points and divide the rotor vibration data sequence into four equal parts; determine the abnormal rotor vibration data interval based on the three rotor vibration data points. N 1, N u ], which belongs to the abnormal rotor vibration data range [ N 1, N u The rotor vibration data was removed to obtain the first rotor vibration dataset; among which, the lower limit of the abnormal rotor vibration data range was... N 1 and upper limit value N u The calculation formulas are as follows: in, n 1 indicates the minimum value of the selected rotor vibration data. n 2 indicates the maximum value of the selected rotor vibration data. ω 1 indicates the first weight. ω 2 indicates the second weight; S12: Generate the rotor vibration matrix based on the first rotor vibration dataset; In step S12, the specific method for generating the rotor vibration matrix is ​​as follows: based on the various rotor vibration data of the first rotor vibration dataset, randomly generate... m × n rotor vibration matrix X Among them, the rotor vibration matrix X Fill in the empty spaces with zeros. m This indicates the row number of the rotor vibration matrix. n Indicates the number of columns in the rotor vibration matrix; S13: Perform a second round of elimination on the rotor vibration matrix to obtain the second rotor vibration dataset; Step S13 includes the following sub-steps: S131: Randomly select a column from the rotor vibration matrix and determine the elimination threshold corresponding to that column; In step S131, the threshold for elimination is... T The calculation formula is: in, x j Represents the rotor vibration matrix X A column vector randomly selected from the data. α Represents a random number between 0 and 1; S132: Remove rotor vibration data that are less than the removal threshold from the column randomly selected in step S131; S133: Repeat step S131 to determine the elimination threshold for each column, and repeat step S132 to eliminate rotor vibration data in each column that are less than the elimination threshold, and use the remaining rotor vibration data as the second rotor vibration dataset. S14: Normalize each vibration data in the second rotor vibration dataset, and use the normalized vibration data as the rotor data sample set. S2: Construct and train a rotor misalignment fault diagnosis model for the rotor data sample set; In step S2, the rotor misalignment fault diagnosis model includes a convolutional layer, a fully connected layer, and an output layer. The convolutional layer is used to extract various features from the rotor data sample set and generate feature differences; the fully connected layer is used to obtain feature vectors based on the feature differences; and the output layer is used to adjust the parameters of the rotor misalignment fault diagnosis model based on the feature vectors. The convolutional layer generates feature differences D ( p The formula for calculating ) is: In the formula, sim (·) represents the similarity function. Y (·) represents the characteristics of the rotor data sample set. Y This indicates the number of features in the rotor data sample set; S3: Use the latest collected rotor vibration data as input to the rotor misalignment fault diagnosis model to perform rotor misalignment fault diagnosis in hydropower stations.

2. The method for diagnosing rotor misalignment faults in hydropower stations according to claim 1, characterized in that, The fully connected layer uses a vector transformation loss function to obtain feature vectors, where the vector transformation loss function... c The expression is: In the formula, M Indicates the number of feature differences. D ( p ) indicates differences in characteristics. α m The mean of the differences in characteristics. β m This represents hyperparameters.

3. The method for diagnosing rotor misalignment faults in hydropower stations according to claim 1, characterized in that, The specific method for adjusting the parameters of the rotor misalignment fault diagnosis model in the output layer is as follows: the softmax function of the output layer is used to classify the feature vectors to obtain the classification probability of each class, and the parameters of the rotor misalignment fault diagnosis model are updated layer by layer according to the classification probability of each class.

4. A hydropower station rotor misalignment fault diagnosis system according to any one of claims 1 to 3, characterized in that, It includes a data acquisition and processing module, a model training module, and a fault diagnosis module connected in sequence; The data acquisition and processing module is used to acquire the initial rotor vibration dataset, preprocess the initial rotor vibration dataset to obtain the rotor data sample set; the model training module is used to construct and train the rotor misalignment fault diagnosis model for the rotor data sample set; the fault diagnosis module is used to use the latest acquired rotor vibration data as the input of the rotor misalignment fault diagnosis model to perform rotor misalignment fault diagnosis in hydropower stations.

5. The hydropower station rotor misalignment fault diagnosis system according to claim 4, characterized in that, The data acquisition and processing module uses several vibration signal acquisition sensors.

6. The hydropower station rotor misalignment fault diagnosis system according to claim 5, characterized in that, Several vibration signal acquisition sensors are fixedly installed on the rotor of the hydropower station to collect vibration data of the rotor and obtain an initial rotor vibration dataset.

7. The hydropower station rotor misalignment fault diagnosis system according to claim 4, characterized in that, The model training module is equipped with a computing power resource submodule, which is used to classify the feature vector using the softmax function of the output layer, obtain the classification probability of each class, and update the parameters of the rotor misalignment fault diagnosis model layer by layer according to the classification probability of each class.

8. An electronic device according to any one of claims 1 to 3, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the hydropower station rotor misalignment fault diagnosis method according to any one of claims 1 to 3.

Images