Water turbine open set fault diagnosis method and device, storage medium and program product
By extracting the historical fault data and category characteristics of the turbine, a target data analysis and category analysis network is built to identify unknown faults, which solves the problem of unidentified unknown faults in the existing technology, and improves the flexibility and accuracy of fault diagnosis.
Patent Information
- Application Number
- CN202410991694.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-23
- Publication Date
- 2024-11-15
AI Technical Summary
Existing deep learning-based turbine fault diagnosis models cannot identify unknown faults, resulting in low flexibility and accuracy in fault identification.
By obtaining the training set, including the historical fault data and categories of the turbine, using the initial data analysis network and category analysis network to extract data characteristics and category characteristics, adjust network parameters, build KL distance loss function, and train the target data analysis network and category analysis network. Identify the target distance between the data to be detected and the preset fault category, distinguishing between known faults and unknown faults.
Improves the flexibility and accuracy of turbine fault diagnosis, can identify unknown fault categories, make full use of known fault data, and ensure high accuracy known category classification performance.
Smart Images

Figure CN118965140A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of hydraulic turbine fault detection, and in particular to a hydraulic turbine open set fault diagnosis method, equipment, storage medium and program product. Background Art
[0002] Hydroelectric generator sets are expensive and are the core and key equipment of hydropower stations. The condition monitoring technology of hydroelectric generator sets can grasp and analyze their operating status at any time, discover existing hidden dangers and defects in time, avoid destructive accidents, and help guide the operation and maintenance of the units, and help ensure the safe and economical operation of the units. It is a necessary means for hydropower plants to achieve "unmanned duty and less manned duty".
[0003] Deep learning technology can automatically extract fault-related features from signals without manual feature engineering. However, most of the current fault diagnosis models based on deep learning are closed-set recognition models, that is, the model classifies all input data into known fault categories. After the model is deployed on site, when encountering unknown faults, the closed-set model will force the unknown faults to be classified as known faults. Therefore, the flexibility and accuracy of fault identification are low. Summary of the invention
[0004] In view of this, the present invention provides a method, device, storage medium and program product for diagnosing open-collection faults of a turbine to solve the problem of being unable to identify unknown faults and improve the accuracy of diagnosing open-collection faults of a turbine.
[0005] In a first aspect, the present invention provides a method for diagnosing an open set fault of a turbine, which comprises: obtaining a training set, the training set comprising historical fault data of an open set of turbines and corresponding historical fault categories; inputting the historical fault data into an initial data analysis network to obtain a data feature distribution of the historical fault data; inputting the historical fault categories into an initial category analysis network to obtain a category feature distribution of the historical fault categories; adjusting the parameters of the initial data analysis network and the initial category analysis network based on the data feature distribution and the category feature distribution to obtain a target data analysis network and a target category analysis network; inputting the data to be detected into the target data analysis network to obtain a target feature distribution; calculating a target distance between the target feature distribution and each preset feature distribution, the preset feature distribution being obtained by processing a plurality of preset fault categories using a target category analysis network; when the minimum target distance is less than or equal to a distance threshold, outputting a fault category of the data to be detected as the preset fault category corresponding to the minimum target distance, and when the minimum target distance is greater than the distance threshold, outputting a fault category of the data to be detected as an unknown fault category.
[0006] In this implementation, the data analysis network and the category analysis network are trained by using the fault data and the fault category, the data to be detected are analyzed by the trained data analysis network, and the known preset fault categories are analyzed by the trained category analysis network, and the fault category closest to the current data to be detected is determined by the target distance between the data to be detected and the preset fault category. The present application does not directly output the closest fault category, but compares the target distance with the distance threshold to distinguish between known faults and unknown faults. Therefore, the method of the present application can not only identify known fault categories, but also identify unknown fault categories, thereby improving the flexibility and accuracy of fault detection.
[0007] In an optional implementation, obtaining a training set includes: obtaining multiple historical fault categories of an open set of turbines; for each historical fault category, collecting multiple historical fault data; and taking each historical fault data and the corresponding historical fault category as a group of data to obtain a training set.
[0008] In an optional embodiment, the initial data analysis network includes an initial shared backbone network, an initial first branch network and an initial second branch network. Inputting historical fault data into the initial data analysis network to obtain data feature distribution of the historical fault data includes: inputting historical fault data into the initial data analysis network, using the initial shared backbone network to perform feature extraction to obtain historical fault feature data; using the initial first branch network to calculate the mean of the historical fault feature data, using the initial second branch network to calculate the variance of the historical fault feature data, and obtaining the data feature distribution of the historical fault data.
[0009] In an optional implementation, the initial category analysis network is an initial prototype Gaussian distribution network, and the historical fault category is input into the initial category analysis network to obtain the category feature distribution of the historical fault category, including: inputting the historical fault category into the initial prototype Gaussian distribution network, calculating the feature space Gaussian distribution of the historical fault category, and obtaining the category feature distribution of the historical fault category.
[0010] In an optional embodiment, the parameters of the initial data analysis network and the initial category analysis network are adjusted based on the data feature distribution and the category feature distribution to obtain the target data analysis network and the target category analysis network, including: calculating the KL distance between each data feature distribution and multiple category feature distributions; constructing a classification loss function based on the KL distance, the classification loss function includes a classification sub-loss function, an intra-class distance sub-loss function and an inter-class distance ion loss function, and the classification loss function is: L = L c +λ1L nd +λ2L ed ; Where L is the classification loss function, L c is the classification loss function, L ndis the intra-class distance loss function, L ed is the inter-class ion loss function, λ1 and λ2 are the sub-loss balance coefficients; based on the classification loss function, the parameters of the initial data analysis network and the initial category analysis network are adjusted to obtain the target data analysis network and the target category analysis network.
[0011] In an optional implementation, constructing a classification loss function based on the KL distance includes: using a SoftMax activation function for the KL distance, and calculating the classification loss function using a cross entropy, the calculation formula is: in, is the KL distance between the i-th data feature distribution and the s-th category feature distribution, y i is the i-th historical fault category, is the KL distance between the data feature distribution of historical fault data and the category feature distribution of the corresponding historical fault category, and γ is a hyperparameter.
[0012] In an optional implementation, the intra-class distance sub-loss function is used to reduce the KL distance between the data feature distribution of the historical fault data and the category feature distribution of the corresponding historical fault category, and constructing a classification loss function based on the KL distance includes: calculating the formula of the intra-class distance sub-loss function is: Among them, y i is the i-th historical fault category, It is the KL distance between the data feature distribution of historical fault data and the category feature distribution of the corresponding historical fault category.
[0013] In an optional implementation, the inter-class ion loss function is used to expand the KL distance between the class feature distributions of each historical fault class, and constructing a classification loss function based on the KL distance includes: calculating the formula for the inter-class ion loss function is: Among them, d s,k is the KL distance between the s-th category feature distribution and the k-th category feature distribution, and m is the hyperparameter of the minimum KL distance between the preset category feature distributions.
[0014] In this implementation, the present application constructs a shared backbone network, two branch networks and a prototype high-speed distribution network, and uses a classification sub-loss function, an intra-class distance sub-loss function and an inter-class distance ion loss function to update the model parameters. The three sub-loss functions use a unified distance metric, which can effectively improve the robustness of feature extraction and improve the accuracy of open set fault detection of the present application.
[0015] In an optional embodiment, before calculating the target distance between the target feature distribution and each preset feature distribution, it also includes: obtaining a cross-validation data set; using the cross-validation data set to determine a distance threshold so that the accuracy of predicting the cross-validation data set as a known fault classification reaches a boundary value of α%, where α is a hyperparameter.
[0016] In this implementation, the present application determines the distance threshold on the basis of ensuring the boundary value of the known class classification accuracy, and can perform unknown fault identification on the basis of ensuring high known class classification performance, making full use of the known fault data on site and improving fault classification accuracy.
[0017] In a second aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the turbine open-collection fault diagnosis method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.
[0018] In a third aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the turbine open-collector fault diagnosis method of the first aspect or any corresponding embodiment thereof.
[0019] In a fourth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for diagnosing open-set faults of a hydraulic turbine according to the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0021] Figure 1 is a flow chart of a method for diagnosing open-collection faults of a hydraulic turbine according to an embodiment of the present invention;
[0022] Figure 2 is a schematic diagram of an open-set fault diagnosis model for a hydraulic turbine according to an embodiment of the present invention;
[0023] Figure 3 is a flow chart of a hydraulic turbine open set fault diagnosis training method according to an embodiment of the present invention;
[0024] Figure 4is a flow chart of another method for diagnosing open-collection faults of a hydraulic turbine according to an embodiment of the present invention;
[0025] Figure 5 is a structural block diagram of a hydraulic turbine open-collector fault diagnosis device according to an embodiment of the present invention;
[0026] Figure 6 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0028] Most of the existing fault diagnosis models based on deep learning are closed set recognition models, that is, the model classifies all input data into known fault categories. Therefore, the fault diagnosis model cannot diagnose unknown faults. Therefore, the present application provides an open set fault diagnosis model for turbines, which can diagnose unknown fault categories on the basis of diagnosing known fault categories. Figure 1 , Figure 1 Schematic diagram of an open-set fault diagnosis model for a hydraulic turbine according to an embodiment of the present invention.
[0029] The open set fault diagnosis model of hydraulic turbine includes data analysis network and category analysis network.
[0030] The data analysis network is used to calculate the characteristic distribution of fault data. The data analysis network includes a shared backbone network, a first branch network and a second branch network.
[0031] The shared backbone network is used to extract features from fault data, and the output data is used as input data for the first branch network and the second branch network respectively.
[0032] The first branch network and the second branch network are parallel networks. The first branch network is used to calculate the mean information of the fault data, and the second branch network is used to calculate the variance information of the fault data.
[0033] In one implementation, the shared backbone network, the first branch network, and the second branch network are all deep neural networks, the shared backbone network contains a parameter θ, and the first branch network contains a parameter The second branch network contains parameter φ.
[0034] Among them, the category analysis network is used to calculate the characteristic distribution of fault categories.
[0035] In one implementation, the category analysis network is a deep neural network. Prototype Gaussian distribution network The prototype Gaussian distribution network is used to calculate the Gaussian distribution of the feature space of the fault category. Where I is the identity matrix.
[0036] In one implementation, the deep neural network includes but is not limited to a convolutional neural network, a fully connected network, a Transformer network, a recurrent neural network, etc.
[0037] In one implementation, the output dimension of the first branch network and the output dimension of the second branch network are μ k The dimension of and the dimension of I are both J.
[0038] Based on this, a method for diagnosing an open-collection fault of a turbine is provided in this embodiment. According to an embodiment of the present invention, an embodiment of a method for diagnosing an open-collection fault of a turbine is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0039] Figure 2 is a flow chart of a method for diagnosing a hydraulic turbine open collector fault according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not based on Figure 2 The process sequence shown is limited.
[0040] like Figure 2 As shown, the process includes the following steps:
[0041] Step S201, obtaining a training set.
[0042] The training set includes the historical fault data of the turbine open set and the corresponding historical fault categories.
[0043] Specifically, the operating status related information of the turbine generating point set under different fault categories is collected as historical fault data. Each historical fault data and its corresponding historical fault category are taken as a group of data, and a training set is constructed using multiple groups of data.
[0044] Step S202, inputting historical fault data into an initial data analysis network to obtain data feature distribution of the historical fault data; inputting historical fault categories into an initial category analysis network to obtain category feature distribution of historical fault categories.
[0045] The training set is output to the initial turbine open set fault diagnosis model, and the historical fault data is analyzed using the initial data analysis network to obtain the data feature distribution of the historical fault data.
[0046] Specifically, the initial shared backbone network is used to perform feature extraction to obtain historical fault feature data, the initial first branch network is used to calculate the mean of the historical fault feature data, the initial second branch network is used to calculate the variance of the historical fault feature data, and the data feature distribution of the historical fault data is obtained by combining the mean and variance.
[0047] The historical fault categories are analyzed using the initial category analysis network to obtain the category feature distribution of the historical fault categories.
[0048] Specifically, the initial prototype Gaussian distribution network is used to calculate the feature space Gaussian distribution of the historical fault categories, and the category feature distribution of the historical fault categories is obtained.
[0049] Step S203, adjusting the parameters of the initial data analysis network and the initial category analysis network based on the data feature distribution and the category feature distribution to obtain the target data analysis network and the target category analysis network.
[0050] Each data feature distribution is compared with multiple category feature distributions to determine the similarity. By adjusting the parameters of the initial data analysis network and the initial category analysis network, the data feature distribution of the historical fault data is less different from the category feature distribution of the corresponding historical fault category, and more different from the category feature distribution of other historical fault categories. After multiple adjustments, the corresponding target data analysis network and target category analysis network are obtained.
[0051] In one implementation, a loss function is constructed using data feature distribution and category feature distribution, and the parameters of the initial data analysis network and the initial category analysis network are reversely adjusted using the loss function.
[0052] In one implementation, the KL distance between each data feature distribution and multiple category feature distributions is calculated, and a classification loss function is constructed based on the KL (Kullback-Leibler) distance.
[0053] Step S204, inputting the data to be detected into the target data analysis network to obtain the target feature distribution.
[0054] The data to be detected is the current operating parameter information of the turbine.
[0055] The trained target data analysis network is used to calculate the Gaussian distribution of the data to be detected and obtain the standard feature distribution.
[0056] Step S205, calculating the target distance between the target feature distribution and each preset feature distribution.
[0057] The preset feature distribution is obtained by processing multiple preset fault categories using the target category analysis network. The preset fault category is a known, trained and tested fault category. The preset fault category includes no fault.
[0058] In one implementation, the KL distance between the target feature distribution and each preset feature distribution is calculated.
[0059] Step S206, when the minimum target distance is less than or equal to the distance threshold, the fault category of the data to be detected is output as the preset fault category corresponding to the minimum target distance; when the minimum target distance is greater than the distance threshold, the fault category of the data to be detected is output as an unknown fault category.
[0060] The multiple target distances are sorted, a preset feature distribution corresponding to the minimum target distance is obtained, and a preset fault category corresponding to the preset feature distribution is further obtained, and the preset fault category is considered to be the most likely current fault category among the multiple preset fault categories.
[0061] Compare the minimum target distance with the distance threshold. When the minimum target distance is less than or equal to the distance threshold, it means that the target feature distribution and the preset feature distribution have a high similarity, and the fault category of the output data to be detected is the preset fault category corresponding to the minimum target distance. When the minimum target distance is greater than the distance threshold, it means that the target feature distribution and the preset feature distribution have a low similarity, and the current fault category does not exist in the currently known preset fault categories, and the fault category of the output data to be detected is an unknown fault category.
[0062] The hydraulic turbine open set fault diagnosis method provided in this embodiment trains a data analysis network and a category analysis network by using fault data and fault categories, analyzes the data to be detected by using the trained data analysis network, and analyzes the known preset fault categories by using the trained category analysis network, and determines the fault category closest to the current data to be detected by the target distance between the data to be detected and the preset fault category. The present application does not directly output the closest fault category, but compares the target distance with the distance threshold to distinguish between known faults and unknown faults. Therefore, the method of the present application can not only identify known fault categories, but also identify unknown fault categories, thereby improving the flexibility and accuracy of fault detection.
[0063] Based on this, a method for training a hydraulic turbine open set fault diagnosis model is provided in this embodiment. Figure 3 is a flow chart of a hydraulic turbine open set fault diagnosis training method according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment does not use Figure 3 The process sequence shown is limited. Figure 3 As shown, the process includes the following steps:
[0064] Step S301, obtaining a training set.
[0065] Specifically, the above step S301 includes:
[0066] Step S3011, collecting operating status related signals of the hydro-generator under different fault categories to obtain historical fault data.
[0067] Specifically, M operation status related signals of the hydro-generator within L time nodes are collected, and the mth operation status related signal is recorded as a m , where a m is a row vector of length L.
[0068] In one implementation, the signals related to the operating status of the hydro-generator include but are not limited to X-direction vibration of the upper frame, Y-direction vibration of the upper frame, Z-direction vibration of the upper frame, X-direction vibration of the lower frame, Y-direction vibration of the lower frame, Z-direction vibration of the lower frame, X-direction vibration of the top cover, Y-direction vibration of the top cover, Z-direction vibration of the top cover, X-direction vibration of the stator frame, Y-direction vibration of the stator frame, Z-direction vibration of the stator frame, vibration of the stator core, swing of the upper and lower guide bearings, swing of the water guide bearing, etc.
[0069] For each row vector a m Perform normalization to obtain the normalized value a ′ m The normalized values of M operation status related signals are stacked to obtain a historical fault data in, The dimension is M×L.
[0070] In one implementation, the normalization calculation method includes maximum normalization, minimum normalization, and Z-score normalization.
[0071] Step S3012: According to the multiple historical fault categories of the turbine, for each historical fault category, multiple historical fault data are collected.
[0072] Specifically, for K different historical fault categories, the method in step S3021 is used to repeatedly collect data on the hydro-generator N times to obtain N historical fault data.
[0073] Step S3013: taking each historical fault data and the corresponding historical fault category as a group of data to obtain a training set.
[0074] Take each historical fault data and the corresponding historical fault category as a group of data to get NK groups of data, and the training set is where xi is the i-th historical fault data, y i is the historical fault data x i The corresponding historical fault category, y i ∈{1,…,K}.
[0075] It can be understood that the N historical fault data x obtained under the same historical fault category correspond to the same historical fault category y.
[0076] Step S302: input the training set into the initial turbine open set fault diagnosis model to obtain data feature distribution and category feature distribution.
[0077] The historical fault data in the training set are input into the data analysis network as training data, and the historical fault data are mapped to the feature space, and the obtained data feature distribution is a Gaussian distribution.
[0078] Specifically, for any historical fault data x i , using the initial first branch network to calculate the mean of the corresponding feature space Gaussian distribution is μ i =E1(E(x i )), the variance of the Gaussian distribution in the corresponding feature space is calculated using the initial second branch network as σ i =E2(E(x i )), the obtained data feature distribution is For NK historical fault data, NK data feature distributions are obtained.
[0079] The historical fault categories in the training set are input into the category analysis network as training labels, and the historical fault categories are mapped to the feature space. The obtained category feature distribution is a Gaussian distribution.
[0080] Specifically, for any historical fault category, the initial prototype Gaussian distribution network is used to calculate the feature space Gaussian distribution to obtain the category feature distribution of the historical fault category. For K historical fault categories, K category feature distributions are obtained.
[0081] Step S303, calculating the KL distance between each data feature distribution and multiple category feature distributions.
[0082] For any historical fault data x i , calculate the KL distance from the data feature distribution to the K category feature distribution, where the calculation formula is:
[0083]
[0084] in, is the KL distance from the i-th data feature distribution to the k-th category feature distribution, symbol *| jrepresents the j-th dimension of *.
[0085] Step S304: construct a classification loss function based on the KL distance.
[0086] Among them, the classification loss function includes the classification sub-loss function, the intra-class distance sub-loss function and the inter-class distance ion loss function.
[0087] A classification sub-loss function is constructed based on the KL distance, and the classification sub-loss function is used to improve the classification accuracy. Specifically, for the KL distance Use the SoftMax activation function to get the historical fault data x i The probability of belonging to the kth historical fault category is:
[0088]
[0089] The classification loss function is calculated using cross entropy:
[0090]
[0091] Among them, γ is a hyperparameter used to control the difficulty of probability assignment.
[0092] Based on the KL distance, an intra-class distance sub-loss function is constructed, which is used to reduce the KL distance between the data feature distribution of the historical fault data and the category feature distribution of the corresponding historical fault category. Specifically, the formula for calculating the intra-class distance sub-loss function is:
[0093]
[0094] Based on the KL distance, a class interval ion loss function is constructed, which is used to expand the KL distance between the class feature distributions of each historical fault category. Specifically, the KL distance between the class feature distribution of the sth historical fault category and the class feature distribution of the kth historical fault category is:
[0095]
[0096] The formula for calculating the inter-class ion loss function is:
[0097]
[0098] Among them, m is the hyperparameter of the minimum KL distance between the preset category feature distributions.
[0099] The classification loss function is calculated using the classification sub-loss function, the intra-class distance sub-loss function, and the inter-class distance ion loss function:
[0100] L=L c +λ1L nd+λ2L ed .
[0101] Among them, λ1 and λ2 are sub-loss balance coefficients.
[0102] The present application constructs a shared backbone network, two branch networks and a prototype distributed network, and uses a classification sub-loss function, an intra-class distance sub-loss function and an inter-class distance ion loss function to update the model parameters. The three sub-loss functions use a unified distance metric, which can effectively improve the robustness of feature extraction and improve the accuracy of open set fault detection in the present application.
[0103] Step S305, adjusting the parameters of the initial data analysis network and the initial category analysis network based on the classification loss function to obtain an open set fault diagnosis model for the target turbine.
[0104] The classification loss function is used to update the model parameters of the initial data analysis network and the initial category analysis network.
[0105] Specifically, update the parameters of the initial shared backbone network Get the target shared backbone network; Update the parameters of the initial first branch network Get the target first branch network; update the parameters of the initial second branch network Get the target second branch network; update the parameters of the initial prototype Gaussian distribution network The target prototype Gaussian distribution network is obtained. Where ν is the learning rate.
[0106] It can be understood that after the target hydraulic turbine open set fault diagnosis model is obtained through training, the target hydraulic turbine open set fault diagnosis model is tested using the test set.
[0107] The hydraulic turbine open set fault diagnosis model training method provided in this embodiment trains the data analysis network and the category analysis network by using the distance relationship between the fault data and the fault category, and updates the model parameters based on the loss function using three unified distance measurement loss functions, thereby improving the robustness of feature extraction and the accuracy of model training after training.
[0108] Based on this, the present application deploys the trained target turbine open set fault diagnosis model to field testing. Therefore, in this embodiment, a turbine open set fault diagnosis method is provided, which can be used for the above-mentioned trained target turbine open set fault diagnosis model. Figure 4 is a flow chart of another method for diagnosing open-collection faults of a hydraulic turbine according to an embodiment of the present invention. It should be noted that if there are substantially the same results, this embodiment is not limited to Figure 4 The process sequence shown is limited. Figure 4 As shown, the process includes the following steps:
[0109] Step S401, input the data to be detected into the target data analysis network to obtain the target feature distribution.
[0110] The data to be tested x t Input the target data analysis network and use the target first branch network to calculate the mean of the Gaussian distribution in the feature space as μ te =E1(E(x t )), the variance of the Gaussian distribution in the feature space is calculated using the target second branch network as σ t =E2(E(x t ), then the target feature distribution is
[0111] Step S402, calculating the KL distance between the target feature distribution and each preset feature distribution.
[0112] Among them, the KL distance between the target feature distribution and the kth preset feature distribution is:
[0113]
[0114] Step S403: obtaining a distance threshold.
[0115] A cross-validation data set is obtained, and a distance threshold is determined using the cross-validation data set so that the accuracy of predicting the cross-validation data set as a known fault classification reaches a boundary value of α%, where α is a hyperparameter.
[0116] The present application determines the distance threshold on the basis of ensuring the boundary value of the known class classification accuracy, and can perform unknown fault identification on the basis of ensuring high known class classification performance, making full use of the known fault data on site and improving the accuracy of fault classification.
[0117] Step S404, determine whether the minimum KL distance is less than or equal to the distance threshold. When the minimum KL distance is less than or equal to the distance threshold, the fault category of the data to be detected is output as the preset fault corresponding to the minimum KL distance. When the minimum KL distance is greater than the distance threshold, the fault category of the data to be detected is output as an unknown fault category.
[0118] The minimum KL distance is determined as the classification result of the target turbine open set fault diagnosis model, where the classification result is:
[0119]
[0120] The minimum KL distance Compare with the distance threshold ε and judge. When the fault category of the output data to be detected is unknown fault category, When , the fault category of the data to be detected is output as the preset fault category corresponding to the minimum KL distance.
[0121] It can be understood that the preset fault category is a known fault category, which includes no fault.
[0122] Furthermore, when the target turbine open-set fault diagnosis model is used to determine that the turbine generator is fault-free, the maintenance strategy is not executed; when the target turbine open-set fault diagnosis model is used to determine that the turbine generator is a known fault, the corresponding maintenance strategy is executed; when the target turbine open-set fault diagnosis model is used to determine that the turbine generator is an unknown fault, the subsequent strategy should be comprehensively determined based on other data.
[0123] The hydraulic turbine open set fault diagnosis method provided in this embodiment analyzes the data to be detected by using a trained data analysis network, analyzes the known preset fault categories by using a trained category analysis network, and determines the fault category closest to the current data to be detected by the target distance between the data to be detected and the preset fault category. The present application does not directly output the closest fault category, but compares the target distance with the distance threshold to distinguish known faults from unknown faults. The distance threshold is determined in advance based on the boundary value of the classification accuracy of the known category. Therefore, the method of the present application can identify unknown fault categories on the basis of identifying known fault categories with high accuracy, thereby improving the flexibility and accuracy of fault detection.
[0124] In this embodiment, a method and device for diagnosing a fault of a hydraulic turbine open collector is also provided, and the device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.
[0125] This embodiment provides a hydraulic turbine open collector fault diagnosis device, Figure 5 is a structural block diagram of a hydraulic turbine open-collector fault diagnosis device according to an embodiment of the present invention. Figure 5 As shown, including:
[0126] The acquisition module 501 is used to acquire a training set, where the training set includes open set historical fault data of the turbine and corresponding historical fault categories.
[0127] The input module 502 is used to input the historical fault data into the initial data analysis network to obtain the data feature distribution of the historical fault data; input the historical fault categories into the initial category analysis network to obtain the category feature distribution of the historical fault categories.
[0128] The training module 503 is used to adjust the parameters of the initial data analysis network and the initial category analysis network based on the data feature distribution and the category feature distribution to obtain the target data analysis network and the target category analysis network.
[0129] The detection module 504 is used to input the data to be detected into the target data analysis network to obtain the target feature distribution.
[0130] The calculation module 505 is used to calculate the target distance between the target feature distribution and each preset feature distribution, where the preset feature distribution is obtained by processing multiple preset fault categories using a target category analysis network.
[0131] The classification module 506 is used to output the fault category of the data to be detected as a preset fault category corresponding to the minimum target distance when the minimum target distance is less than or equal to the distance threshold, and to output the fault category of the data to be detected as an unknown fault category when the minimum target distance is greater than the distance threshold.
[0132] The further functional description of each of the above modules is the same as that of the above corresponding embodiments and will not be repeated here.
[0133] The turbine open-collector fault diagnosis device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.
[0134] The embodiment of the present invention also provides a computer device having the above Figure 5 The turbine open-collector fault diagnosis device shown.
[0135] See also Figure 6 , Figure 6 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 6As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 6 A processor 10 is taken as an example.
[0136] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0137] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.
[0138] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0139] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0140] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.
[0141] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.
[0142] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0143] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of computer program instructions in computer-readable media includes, but is not limited to, source files, executable files, installation package files, etc., and accordingly, the way in which computer program instructions are executed by a computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.
[0144] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A method for diagnosing open-collection faults of a hydraulic turbine, characterized in that: The method comprises: Acquire a training set, wherein the training set includes historical fault data of an open set of hydraulic turbines and corresponding historical fault categories; Inputting the historical fault data into an initial data analysis network to obtain data feature distribution of the historical fault data; inputting the historical fault categories into an initial category analysis network to obtain category feature distribution of the historical fault categories; Adjusting the parameters of the initial data analysis network and the initial category analysis network based on the data feature distribution and the category feature distribution to obtain a target data analysis network and a target category analysis network; Inputting the data to be detected into the target data analysis network to obtain target feature distribution; Calculating a target distance between the target feature distribution and each preset feature distribution, wherein the preset feature distribution is obtained by processing a plurality of preset fault categories using the target category analysis network; When the minimum target distance is less than or equal to the distance threshold, the fault category of the data to be detected is the preset fault category corresponding to the minimum target distance; when the minimum target distance is greater than the distance threshold, the fault category of the data to be detected is an unknown fault category.
2. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 1, characterized in that: The obtaining of the training set comprises: Acquire a plurality of the historical fault categories of the turbine open set; For each of the historical fault categories, collecting a plurality of the historical fault data; Each of the historical fault data and the corresponding historical fault category is taken as a group of data to obtain the training set.
3. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 2, characterized in that: The initial data analysis network includes an initial shared backbone network, an initial first branch network, and an initial second branch network. Inputting the historical fault data into the initial data analysis network to obtain the data feature distribution of the historical fault data includes: Inputting the historical fault data into the initial data analysis network, and performing feature extraction using the initial shared backbone network to obtain historical fault feature data; The mean of the historical fault feature data is calculated using the initial first branch network, and the variance of the historical fault feature data is calculated using the initial second branch network to obtain the data feature distribution of the historical fault data.
4. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 3, characterized in that: The initial category analysis network is an initial prototype Gaussian distribution network, and the inputting of the historical fault category into the initial category analysis network to obtain the category feature distribution of the historical fault category includes: The historical fault categories are input into the initial prototype Gaussian distribution network, the feature space Gaussian distribution of the historical fault categories is calculated, and the category feature distribution of the historical fault categories is obtained.
5. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 4, characterized in that: The adjusting the parameters of the initial data analysis network and the initial category analysis network based on the data feature distribution and the category feature distribution to obtain a target data analysis network and a target category analysis network comprises: Calculating the KL distance between each of the data feature distributions and a plurality of the category feature distributions; A classification loss function is constructed based on the KL distance. The classification loss function includes a classification sub-loss function, an intra-class distance sub-loss function, and an inter-class distance ion loss function. The classification loss function is: L=L c +λ1L nd +λ2L ed ; Wherein, L is the classification loss function, and L c is the classifier loss function, the L nd is the intra-class distance loss function, the L ed is the inter-class ion loss function, and the λ1 and λ2 are sub-loss balance coefficients; The parameters of the initial data analysis network and the initial category analysis network are adjusted based on the classification loss function to obtain the target data analysis network and the target category analysis network.
6. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 5, characterized in that: The constructing of a classification loss function based on the KL distance comprises: The SoftMax activation function is used for the KL distance, and the classification loss function is calculated using the cross entropy. The calculation formula is: Among them, the is the KL distance between the i-th data feature distribution and the s-th category feature distribution, the y i is the i-th historical fault category, is the KL distance between the data feature distribution of the historical fault data and the corresponding category feature distribution of the historical fault category, and γ is a hyperparameter.
7. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 5, characterized in that: The intra-class distance sub-loss function is used to reduce the KL distance between the data feature distribution of the historical fault data and the category feature distribution of the corresponding historical fault category. The classification loss function constructed based on the KL distance includes: The formula for calculating the intra-class distance sub-loss function is: Among them, the y i is the i-th historical fault category, It is the KL distance between the data feature distribution of the historical fault data and the corresponding category feature distribution of the historical fault category.
8. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 5, characterized in that: The inter-class loss function is used to expand the KL distance between the category feature distributions of each of the historical fault categories, and constructing a classification loss function based on the KL distance includes: The formula for calculating the inter-class ion loss function is: Among them, the d s , is the KL distance between the s-th category feature distribution and the k-th category feature distribution, and m is a hyperparameter of the preset minimum KL distance between the category feature distributions.
9. The method for diagnosing open-collection faults of a hydraulic turbine according to claim 5, characterized in that: Before calculating the target distance between the target feature distribution and each of the preset feature distributions, the method further includes: Get the cross validation dataset; The distance threshold is determined using the cross-validation data set so that the accuracy of predicting the cross-validation data set as a known fault classification reaches a boundary value of α%, wherein α is a hyperparameter.
10. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the turbine open-collector fault diagnosis method according to any one of claims 1 to 9 by executing the computer instructions.
11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the turbine open-collector fault diagnosis method according to any one of claims 1 to 9.
12. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to enable a computer to execute the hydraulic turbine open-collector fault diagnosis method according to any one of claims 1 to 9.
Citation Information
Cited By
Water turbine open-set fault diagnosis method, device, storage medium, and program product
WO2026020733A1
