A clustering-guided partial migration fault diagnosis method for hydro-generator units based on unsupervised domain adversarial network

Abstract

The present invention is a method for diagnosing partial migration faults of hydro-turbine generator sets based on an unsupervised domain adversarial network under clustering guidance. An unsupervised domain adversarial network based on a batch normalization method for a specific domain is constructed to remove specific domain information and enhance the alignment between source and target domain features; an embedded clustering strategy is designed to learn tightly clustered target domain features during partial domain adaptation to suppress the negative migration phenomenon during partial domain adaptation; a joint optimization objective function is defined to balance different loss functions. This method effectively solves the problem that the target domain features extracted by the existing unsupervised partial domain adaptation model during iterative training are relatively scattered, which easily leads to negative migration. The proposed method can accurately extract transferable features between the source and target domains, suppress the negative migration phenomenon during partial domain adaptation, and achieve higher diagnostic accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of hydro-generator fault detection, and in particular to a cluster-guided hydro-generator set partial migration fault diagnosis method based on an unsupervised domain adversarial network. Background Art

[0002] In recent years, with the rapid advancement of modern industry, the reliability and safety requirements for integrated machine systems in hydropower plants have increased significantly. After long-term operation, critical rotating components of hydropower generators, such as bearings and runners, face complex and diverse failure challenges. These failures not only erode equipment health but can also cause serious safety incidents, posing a significant threat to power plant operations. Therefore, condition monitoring and accurate fault diagnosis of hydropower generators have become an indispensable and critical component of power plant operation and maintenance management systems. Relying on advanced real-time online monitoring systems and vibration monitoring technologies, we are able to capture and analyze massive amounts of operational data in real time. We have developed data-driven and model-driven fault diagnosis technologies, providing solid data support and a scientific basis for the formulation of condition-based maintenance strategies. Through early warning and immediate intervention, we effectively curb the deterioration of potential faults, significantly reducing the risk of major operational accidents and building a solid defense for the safe, stable, and efficient operation of hydropower plants.

[0003] Existing research methods mainly focus on fault diagnosis of hydro-turbine generator sets based on feature extraction methods. The accuracy of fault prediction is improved by extracting feature information from monitoring signals that can fully characterize the operating status of the unit, or by constructing intelligent fault diagnosis models based on methods such as convolutional neural networks, which automatically extract high-dimensional features directly from the original monitoring data and reduce the reliance on manual feature design. These networks can learn complex patterns in data and improve the accuracy of fault identification, but there are still some problems that need to be improved: (1) In actual industrial scenarios, since hydro-turbine generator sets are in a healthy state for a long time, the number of fault state samples is very limited, which makes it difficult to meet the assumption that deep learning requires a large number of labeled and uniformly distributed fault samples; (2) Data-driven intelligent fault diagnosis models need to collect complete data on various fault types of target mechanical equipment, but in some cases, it is impossible to collect data sets for all fault states, making the category space of the target domain only a subset of the category space of the source domain. The domain adaptation effect of the model will be greatly reduced, which is prone to negative transfer problems.

[0004] Therefore, new technologies need to be introduced to accurately diagnose the status of the unit, achieve precise monitoring of the unit's operating status and fault warning, and provide strong guarantees for the safe and stable operation of the hydropower station. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a partial migration fault diagnosis method for hydro-generator sets based on an unsupervised domain adversarial network under clustering guidance, so as to realize accurate monitoring of the operating status of the set and fault warning, and provide strong guarantee for the safe and stable operation of the hydropower station.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: a cluster-guided hydro-generator partial migration fault diagnosis method based on an unsupervised domain adversarial network, comprising the following steps:

[0007] Step S1: collect vibration data of hydro-generator sets with similar structures and parameters to actual equipment in the project to construct a source domain dataset, and divide the source domain into training samples and test samples;

[0008] Step S2: collecting vibration data of a hydro-generator set measured in a power plant to construct a target domain dataset, and dividing the target domain into training samples and test samples;

[0009] Step S3: linearly transform the source domain and target domain data to the range of [0, 1] using the minimum-maximum normalization method;

[0010] Step S4: construct an unsupervised domain adversarial network based on domain-specific batch normalization, apply the DSBN layer to remove domain-specific interference information in the network model, enhance the alignment of source and target domain feature distributions to extract transferable features;

[0011] Step S5: Using an embedded clustering strategy based on KL divergence, during the partial domain adaptation process, each target domain feature is clustered into one of the source domain categories, and the model is trained and optimized.

[0012] Step S6, a joint optimization objective function is defined to balance different losses to improve training and diagnosis performance, including embedded clustering loss, weighted domain classification loss, and weighted domain adaptation adversarial loss;

[0013] Step S7: Input the training samples of the source domain and the target domain, use the Adam algorithm to iteratively train and optimize the proposed network, and obtain the partial migration fault diagnosis model of the hydro-generator set through training. Input the target domain test samples into the trained network for verification;

[0014] Step S8: using the trained hydro-generator set partial migration fault diagnosis model to identify the operating status of the hydro-generator set.

[0015] Preferably, step S4 includes the following steps:

[0016] Step S41: Applying the domain-specific batch normalization method DSBN (DSBN) instead of the batch normalization method BN (BN) to construct an unsupervised domain adversarial network, which includes two independent BN branches specifically for source domain and target domain samples. During training, each sample selects its corresponding BN branch in the DSBN according to its corresponding domain; represents the input sample of the domain adversarial network, where , represents a specific set of parameters for the source domain, Representing the specific parameter set of the target domain, the DSBN layer is represented as;

[0017] ;

[0018] in:

[0019] ;

[0020] in, and represent In the model training phase, DSBN first divides the source and target samples into different small batches, and then calculates the mean and variance of the source and target domains respectively. In the model testing phase, the mean and variance of each domain are used for the samples in the corresponding domain to perform batch normalization operations.

[0021] Preferably, step S5 includes the following steps:

[0022] Step S51: A target domain augmentation strategy is designed. This strategy uses the k-means method to determine target domain features and expands some source domain samples into the target domain category space to fix the number of clusters. For example, during training, 60% of each batch is randomly sampled as target domain samples and 40% as source domain samples (without labels). This ensures that the augmented target domain category space is identical to the source domain category space, allowing the partial domain adaptation problem to naturally transition to a pseudo-domain adaptation problem, with the number of clusters equal to the number of source classes. Furthermore, the embedded clustering strategy makes the learned source and target domain sample features more difficult to distinguish, resulting in a softer boundary for the target set trained by the domain discriminator and helping to reduce the impact of negative transfer.

[0023] Step S52: Select the KL divergence between the soft distribution and the auxiliary distribution as the clustering objective function to train and optimize the proposed model. The clustering objective function is expressed as:

[0024] ;

[0025] in, is the clustering objective function, and represent soft allocation and auxiliary target distribution respectively; The target domain sample Divided into categories The probability of , that is, soft allocation; Distribute probabilities for predefined high-confidence target domains;

[0026] Step S53, in the calculation When using −Distribution is measured as an embedded kernel function and The closer to the cluster centroid , probability The higher:

[0027] ;

[0028] ;

[0029] in, for − degrees of freedom of the distribution, For the Euclidean distance metric, calculate each target domain feature and similarity;

[0030] Step S54: In order to standardize the loss contribution of each centroid and prevent large clusters from distorting the hidden feature space, first Raised to the power of 2, and then normalized according to the frequency of each cluster, to calculate :

[0031] ;

[0032] in, Indicates the frequency of soft clusters.

[0033] Preferably, the step S6 is specifically as follows:

[0034] The joint optimization objective function is expressed as:

[0035] ;

[0036] ;

[0037] ;

[0038] ;

[0039] ;

[0040] in, To jointly optimize the objective function, is the embedded clustering loss function, represents the weighted classification loss, represents the weighted domain adaptive adversarial loss, and It is a hyperparameter used to balance different losses in the joint optimization objective function. Indicates the number of source domain clusters, is the indicator function, and represents the feature generator and feature classifier, represents the partially adversarial domain discriminator, The target sample is processed by the Softmax function The predicted label value of yes The dimension weight vector represents the contribution weight of each source domain category; is a source domain sample The category weights, is a constant. As the number of iterations increases, From 0 to 1, It increases from 0 to 1. It is worth noting that in the initial training stage, the clustering loss Fixed to 0. When the number of training cycles reaches the preset value, is activated and set to a non-negative constant.

[0041] The beneficial effects of the clustering-guided method for diagnosing partial migration faults of hydro-turbine generator sets based on an unsupervised domain adversarial network provided by the present invention are as follows: Step S4, constructing an unsupervised domain adversarial network based on specific domain batch normalization, applying the DSBN layer to remove interference information of the specific domain in the network model, and enhancing the alignment of the source domain and target domain feature distributions to extract transferable features; Step S5, adopting an embedded clustering strategy based on KL divergence to learn tightly clustered target domain features during the partial domain adaptation process, and suppress the negative transfer phenomenon during the partial domain adaptation process; Step S6, defining a joint optimization objective function to balance different loss functions to improve the fault diagnosis accuracy of key components of rotating machinery. The present invention solves the problem that the target domain features extracted by the existing unsupervised partial domain adaptation model during iterative training are relatively scattered, which easily leads to negative transfer. The method of the present invention enables the model to learn tightly clustered target domain features during the partial domain adaptation process, suppress the negative transfer phenomenon, and achieve higher diagnostic accuracy in partial migration faults of hydro-turbine generator sets. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention will be further described below with reference to the accompanying drawings and examples:

[0043] Figure 1 is a flow chart of the method of the present invention;

[0044] Figure 2 Schematic diagram of the construction of the unsupervised domain adversarial network of the present invention. DETAILED DESCRIPTION

[0045] A cluster-guided partial migration fault diagnosis method for a hydro-generator set based on an unsupervised domain adversarial network comprises the following steps:

[0046] Step S1: collect vibration data of hydro-generator sets with similar structures and parameters to actual equipment in the project to construct a source domain dataset, and divide the source domain into training samples and test samples;

[0047] Step S2: collecting vibration data of a hydro-generator set measured in a power plant to construct a target domain dataset, and dividing the target domain into training samples and test samples;

[0048] Step S3: linearly transform the source domain and target domain data to the range of [0, 1] using the minimum-maximum normalization method;

[0049] Step S4: construct an unsupervised domain adversarial network based on domain-specific batch normalization, apply the DSBN layer to remove domain-specific interference information in the network model, enhance the alignment of source and target domain feature distributions to extract transferable features;

[0050] Step S5: Using an embedded clustering strategy based on KL divergence, during the partial domain adaptation process, each target domain feature is clustered into one of the source domain categories, and the model is trained and optimized.

[0051] Step S6, a joint optimization objective function is defined to balance different losses to improve training and diagnosis performance, including embedded clustering loss, weighted domain classification loss, and weighted domain adaptation adversarial loss;

[0052] Step S7: Input the training samples of the source domain and the target domain, use the Adam algorithm to iteratively train and optimize the proposed network, and obtain the partial migration fault diagnosis model of the hydro-generator set through training. Input the target domain test samples into the trained network for verification;

[0053] Step S8: using the trained hydro-generator set partial migration fault diagnosis model to identify the operating status of the hydro-generator set.

[0054] Preferably, step S4 includes the following steps:

[0055] Step S41: Applying the domain-specific batch normalization method DSBN (DSBN) instead of the batch normalization method BN (BN) to construct an unsupervised domain adversarial network, which includes two independent BN branches specifically for source domain and target domain samples. During training, each sample selects its corresponding BN branch in the DSBN according to its corresponding domain; represents the input sample of the domain adversarial network, where , represents a specific set of parameters for the source domain, Representing the specific parameter set of the target domain, the DSBN layer is represented as;

[0056] ;

[0057] in:

[0058] ;

[0059] in, and represent In the model training phase, DSBN first divides the source and target samples into different small batches, and then calculates the mean and variance of the source and target domains respectively. In the model testing phase, the mean and variance of each domain are used for the samples in the corresponding domain to perform batch normalization operations.

[0060] Preferably, step S5 includes the following steps:

[0061] Step S51: A target domain augmentation strategy is designed. This strategy uses the k-means method to determine target domain features and expands some source domain samples into the target domain category space to fix the number of clusters. For example, during training, 60% of each batch is randomly sampled as target domain samples and 40% as source domain samples (without labels). This ensures that the augmented target domain category space is identical to the source domain category space, allowing the partial domain adaptation problem to naturally transition to a pseudo-domain adaptation problem, with the number of clusters equal to the number of source classes. Furthermore, the embedded clustering strategy makes the learned source and target domain sample features more difficult to distinguish, resulting in a softer boundary for the target set trained by the domain discriminator and helping to reduce the impact of negative transfer.

[0062] Step S52: Select the KL divergence between the soft distribution and the auxiliary distribution as the clustering objective function to train and optimize the proposed model. The clustering objective function is expressed as:

[0063] ;

[0064] in, is the clustering objective function, and represent soft allocation and auxiliary target distribution respectively; The target domain sample Divided into categories The probability of , that is, soft allocation; Distribute probabilities for predefined high-confidence target domains;

[0065] Step S53, in the calculation When using −Distribution is measured as an embedded kernel function and The closer to the cluster centroid , probability The higher:

[0066] ;

[0067] ;

[0068] in, for − degrees of freedom of the distribution, For the Euclidean distance metric, calculate each target domain feature and similarity;

[0069] Step S54: In order to standardize the loss contribution of each centroid and prevent large clusters from distorting the hidden feature space, first Raised to the power of 2, and then normalized according to the frequency of each cluster, to calculate :

[0070] ;

[0071] in, Indicates the frequency of soft clusters.

[0072] Preferably, the step S6 is specifically as follows:

[0073] The joint optimization objective function is expressed as:

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] ;

[0079] in, To jointly optimize the objective function, is the embedded clustering loss function, represents the weighted classification loss, represents the weighted domain adaptive adversarial loss, and It is a hyperparameter used to balance different losses in the joint optimization objective function. Indicates the number of source domain clusters, is the indicator function, and represents the feature generator and feature classifier, represents the partially adversarial domain discriminator, The target sample is processed by the Softmax function The predicted label value of yes The dimension weight vector represents the contribution weight of each source domain category; is a source domain sample The category weights, is a constant. As the number of iterations increases, From 0 to 1, It increases from 0 to 1. It is worth noting that in the initial training stage, the clustering loss Fixed to 0. When the number of training cycles reaches the preset value, is activated and set to a non-negative constant.

[0080] In this example, the source domain dataset was constructed using vibration data from a hydro-turbine generator set with similar structure and parameters to actual engineering equipment. The target domain dataset was constructed using vibration data collected from a hydro-turbine generator set at a power plant. The collected vibration acceleration signals included normal conditions, bearing inner race faults, bearing outer race faults, bearing rolling element faults, bearing retainer faults, and insulation faults, as shown in Table 1. For each fault mode, 200 samples were extracted, of which 160 (80%) were used as training samples, and the remaining were used as testing samples.

[0081] Table 1 Detailed information of the diagnostic dataset

[0082] Fault Label Failure Mode 1 Normal state 2 Bearing inner ring failure 3 Bearing outer ring failure 4 Bearing rolling element failure 5 Bearing cage failure 6 Insulation failure

[0083] The network structure settings are shown in Table 2, and the training parameters are as follows: batch The value range of is [64,256], and the value range of the initial learning rate is {10 -4 , 10 -3 , 10-2 , 10 -1}, the number of iterations is 300, the preset number of iterations is 20, and the Adam optimization algorithm is selected for training. In addition, it is found through experience that the preset number of iterations set to 20 can provide a good initialization point for the embedded clustering loss.

[0084] Table 2 Domain adversarial network structure and parameters

[0085]

[0086] To validate the effectiveness of the proposed method, four different partial transfer learning tasks were set up. Details of the settings are shown in Table 3. To reduce random errors, each partial transfer learning task was repeated 10 times. Several popular transfer learning methods were evaluated and compared, including CNNs, Gaussian kernel-induced MMD domain adaptation, partial adversarial domain adaptation, and importance-weighted adversarial networks.

[0087] Table 3 Experimental setup information for some transfer learning tasks

[0088] Task Name Source domain category space Target domain category space A1 1,2,3,4,5,6 1,2,3,4,5,6 A2 1,2,3,4,5,6 1,2,3,4,5 A3 1,2,3,4,5,6 1,2,3,4 A4 1,2,3,4,5,6 1,2,3

[0089] Table 4 shows the diagnostic accuracy of different methods in partial transfer learning tasks, where L represents the diagnostic accuracy of the method in the last iteration. In addition, to quantitatively evaluate the impact of negative transfer effects, Table 4 also lists the highest diagnostic accuracy of the method throughout the training process (denoted as M). The table shows that compared with the other four popular transfer learning methods, the proposed method achieves superior diagnostic accuracy on both classic transfer learning tasks (A1) and partial transfer learning tasks (A2, A3, and A4). This shows that the embedded clustering strategy can suppress negative transfer in the local domain adaptation process, and the model's domain-specific batch normalization method can strengthen the alignment between the two domains to a certain extent.

[0090] Table 4 Diagnostic accuracy of different methods in some transfer learning tasks

[0091]

[0092] like Figure 1As shown, the content of the present invention can be mainly divided into four parts. The first part is to collect the vibration data of the turbine generator set that is relatively similar to the actual equipment structure and parameters of the engineering project to construct a source domain data set, and to collect the vibration data of the turbine generator set from the measured turbine generator set of a certain power plant to construct a target domain data set and divide the target domain training samples and test samples. After de-meaning and normalization, they are divided into training samples and test samples of the source domain and target domain respectively; the second part is to construct an unsupervised domain adversarial network based on specific domain batch normalization, apply the DSBN layer to remove the interference information of the specific domain in the network model, enhance the alignment of the source domain and target domain feature distribution to extract transferable features; the third part is to adopt an embedded clustering strategy based on KL divergence, and promote each target domain feature to be clustered into a category in the source domain category during the partial domain adaptation process, and train and optimize the model; the fourth part is to define a joint optimization objective function to balance different losses to improve training and diagnostic performance; the fifth part is to use the Adam algorithm to iteratively train and optimize the proposed network, and obtain a partial migration fault diagnosis model of the turbine generator set through training, and input the target domain test sample into the trained network for verification;

[0093] like Figure 2 As shown in Figure 3, the basic structural modules of the proposed network model include feature generators, domain discriminators, and classifiers. DSBN contains two independent BN branches specifically for source and target domain samples. During training, each sample selects its corresponding BN branch in DSBN according to its corresponding domain.

[0094] The present invention is a method for diagnosing partial migration faults of hydro-turbine generator sets based on an unsupervised domain adversarial network under clustering guidance. An unsupervised domain adversarial network based on a batch normalization method for a specific domain is constructed to remove specific domain information and enhance the alignment between source and target domain features; an embedded clustering strategy is designed to learn tightly clustered target domain features during partial domain adaptation to suppress the negative migration phenomenon during partial domain adaptation; a joint optimization objective function is defined to balance different loss functions. This method effectively solves the problem that the target domain features extracted by the existing unsupervised partial domain adaptation model during iterative training are relatively scattered, which easily leads to negative migration. The proposed method can accurately extract transferable features between the source domain and the target domain, suppress the negative migration phenomenon during partial domain adaptation, and achieve higher diagnostic accuracy, demonstrating its effectiveness and superiority in solving the problem of unsupervised partial migration fault diagnosis of hydro-turbine generator sets, and is superior to the existing technology.

[0095] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A cluster-guided partial migration fault diagnosis method for hydro-generator units based on an unsupervised domain adversarial network, characterized in that: The following steps are involved: Step S1: collect vibration data of hydro-generator sets with similar structures and parameters to actual equipment in the project to construct a source domain dataset, and divide the source domain into training samples and test samples; Step S2: collecting vibration data of a hydro-generator set measured in a power plant to construct a target domain dataset, and dividing the target domain into training samples and test samples; Step S3: linearly transform the source domain and target domain data to the range of [0, 1] using the minimum-maximum normalization method; Step S4: construct an unsupervised domain adversarial network based on domain-specific batch normalization, apply the DSBN layer to remove domain-specific interference information in the network model, enhance the alignment of source and target domain feature distributions to extract transferable features; Step S5: Using an embedded clustering strategy based on KL divergence, during the partial domain adaptation process, each target domain feature is clustered into one of the source domain categories, and the model is trained and optimized. Step S6, a joint optimization objective function is defined to balance different losses to improve training and diagnosis performance, including embedded clustering loss, weighted domain classification loss, and weighted domain adaptation adversarial loss; Step S7: Input the training samples of the source domain and the target domain, use the Adam algorithm to iteratively train and optimize the proposed network, and obtain the partial migration fault diagnosis model of the hydro-generator set through training. Input the target domain test samples into the trained network for verification; Step S8: using the trained hydro-generator set partial migration fault diagnosis model to identify the operating status of the hydro-generator set.

2. According to claim 1, a cluster-guided hydro-generator partial migration fault diagnosis method based on an unsupervised domain adversarial network is characterized in that: The step S4 comprises the following steps: Step S41: Apply a domain-specific batch normalization method (DSBN) instead of a batch normalization method (BN) to construct an unsupervised domain adversarial network, which includes two independent BN branches specifically for source and target domain samples. During training, each sample selects its corresponding BN branch in the DSBN according to its corresponding domain. represents the input sample of the domain adversarial network, where , represents a specific set of parameters for the source domain, Representing a specific parameter set for the target domain, the DSBN layer is represented as: ； in: ； in, and represent In the model training phase, DSBN first divides the source and target samples into different small batches, and then calculates the mean and variance of the source and target domains respectively. In the model testing phase, the mean and variance of each domain are used for the samples in the corresponding domain to perform batch normalization operations.

3. The method for diagnosing partial migration faults of hydro-generator sets based on unsupervised domain adversarial networks under clustering guidance according to claim 1 is characterized in that: The step S5 comprises the following steps: Step S51: Use the k-means method to determine the target domain features, and expand some source domain samples into the target domain category space to fix the value of the number of clusters; Step S52: Select the KL divergence between the soft distribution and the auxiliary distribution as the clustering objective function to train and optimize the proposed model. The clustering objective function is expressed as: ； in, is the clustering objective function, and represent soft allocation and auxiliary target distribution respectively; The target domain sample Divided into categories The probability of , that is, soft allocation; Distribute probabilities for predefined high-confidence target domains; Step S53, in the calculation When using −Distribution is measured as an embedded kernel function and The closer to the cluster centroid , probability The higher: ； ； in, for − degrees of freedom of the distribution, For the Euclidean distance metric, calculate each target domain feature and similarity; Step S54: In order to standardize the loss contribution of each centroid and prevent large clusters from distorting the hidden feature space, first Raised to the power of 2, and then normalized according to the frequency of each cluster, to calculate : ； in, Indicates the frequency of soft clusters.

4. The method for diagnosing partial migration faults of hydro-generator sets based on unsupervised domain adversarial networks under clustering guidance according to claim 1 is characterized in that: The step S6 is specifically as follows: The joint optimization objective function is expressed as: ； ； ； ； ； in, To jointly optimize the objective function, is the embedded clustering loss function, represents the weighted classification loss, represents the weighted domain adaptive adversarial loss, and It is a hyperparameter used to balance different losses in the joint optimization objective function. Indicates the number of source domain clusters, is the indicator function, and represents the feature generator and feature classifier, represents the partially adversarial domain discriminator, The target sample is processed by the Softmax function The predicted label value of yes The dimension weight vector represents the contribution weight of each source domain category; is a source domain sample The category weights, is a constant. As the number of iterations increases, From 0 to 1, Increase from 0 to 1.

Images