Guide vane detection method and device, storage medium, electronic equipment and program product

Abstract

The invention relates to a guide vane detection method and device, a storage medium, electronic equipment and a program product, and relates to the technical field of guide vanes, and the method can comprise the following steps: obtaining guide vane parameters of a guide vane at a target moment; determining a guide vane feature vector corresponding to the guide vane blade parameter; inputting the guide vane feature vector into a pre-constructed target detection model to obtain a guide vane operation state result output by the target detection model; the operation state result is used for representing that the guide vane is in a normal operation state or an abnormal operation state. Therefore, the current operation state is analyzed in real time through the forward propagation and adversarial generation mechanism of the target detection model, so that the operation state result of the guide vane is accurately identified, the real-time performance and accuracy of abnormal operation detection of the guide vane are ensured, the accuracy and reliability of abnormal detection are greatly improved, and the situations of missing report and false report are reduced.

Description

Technical Field

[0001] This disclosure relates to the field of guide vane technology, and more specifically, to a method, apparatus, storage medium, electronic device, and program product for detecting guide vanes. Background Technology

[0002] Guide vanes are key components in water turbines or gas turbines, used to regulate fluid flow and optimize equipment performance. Their proper operation is crucial to the overall efficiency and safety of the turbine. However, during long-term operation, guide vanes are prone to failure due to mechanical wear, hydraulic system malfunctions, or environmental changes, thus affecting the operation of the water turbine and gas turbine. Summary of the Invention

[0003] To address the aforementioned technical problems, this disclosure provides a method, apparatus, storage medium, electronic device, and program product for detecting guide vanes.

[0004] In a first aspect, this disclosure provides a method for detecting guide vanes, the method comprising: Obtain the guide vane blade parameters at the target time; Determine the guide vane feature vector corresponding to the guide vane blade parameters; The guide vane feature vector is input into a pre-constructed target detection model to obtain the operating status result of the guide vane output by the target detection model; the operating status result is used to characterize whether the guide vane is in a normal operating state or an abnormal operating state. The target detection model consists of an encoder and a generative adversarial network. The encoder is used to generate corresponding encoded features based on the guide vane feature vector, and the generative adversarial network is used to predict the encoded features and generate corresponding running status results.

[0005] Optionally, determining the guide vane feature vector corresponding to the guide vane blade parameters includes: The guide vane blade parameters are pre-processed to obtain target guide vane parameters; the pre-processing includes one or more of the following processing methods: noise reduction processing, normalization processing, and deletion of abnormal data; Determine the guide vane feature vector corresponding to the target guide vane parameters.

[0006] Optionally, the generative adversarial network is pre-trained in the following manner: Obtain a preset training sample set, which includes multiple sample guide vane feature vectors and a running status label corresponding to each sample guide vane feature vector; the running status label includes normal running status or abnormal running status. The preset training model is trained based on the multiple sample guide vane feature vectors and the running state label corresponding to each sample guide vane feature vector, and the trained preset training model is used as the generative adversarial network.

[0007] Optionally, the method further includes: If the operating status result includes an abnormal operating status, obtain the historical guide vane parameters within the historical time period prior to the target time. Based on the historical guide vane parameters and the guide vane blade parameters, determine the cause of the guide vane failure.

[0008] Optionally, determining the cause of the guide vane failure based on the historical guide vane parameters and the guide vane blade parameters includes: Determine the parameter difference between each of the historical guide vane parameters and the guide vane blade parameters; Based on the parameter differences, the historical guide vane parameters and the guide vane blade parameters are clustered to obtain multiple cluster sets; Based on the multiple cluster sets, the cause of the failure corresponding to the guide vane is determined.

[0009] Optionally, the method further includes: Based on the historical guide vane parameters, the guide vane blade parameters, and the cause of failure, determine the target probability of the guide vane experiencing the cause of failure in the future time. Based on the target probability, a corresponding guide vane maintenance scheme is generated.

[0010] Secondly, this disclosure provides a detection device for a guide vane, the device comprising: The acquisition module is used to acquire the guide vane blade parameters at the target time. The determination module is used to determine the guide vane feature vector corresponding to the guide vane blade parameters; The detection module is used to input the feature vector of the guide vane into a pre-constructed target detection model to obtain the operating status result of the guide vane output by the target detection model; the operating status result is used to characterize whether the guide vane is in a normal operating state or an abnormal operating state. The target detection model consists of an encoder and a generative adversarial network. The encoder is used to generate corresponding encoded features based on the guide vane feature vector, and the generative adversarial network is used to predict the encoded features and generate corresponding running status results.

[0011] Optionally, the determining module is used to perform preset processing on the guide vane blade parameters to obtain target guide vane parameters; the preset processing includes one or more of the following processing methods: noise reduction processing, normalization processing, and deletion of abnormal data; and to determine the guide vane feature vector corresponding to the target guide vane parameters.

[0012] Optionally, the generative adversarial network is pre-trained in the following manner: Obtain a preset training sample set, which includes multiple sample guide vane feature vectors and a running status label corresponding to each sample guide vane feature vector; the running status label includes normal running status or abnormal running status. The preset training model is trained based on the multiple sample guide vane feature vectors and the running state label corresponding to each sample guide vane feature vector, and the trained preset training model is used as the generative adversarial network.

[0013] Optionally, the acquisition module is further configured to acquire historical guide vane parameters within a historical time period prior to the target time if the operating status result includes an abnormal operating status. The determining module is further configured to determine the cause of the fault corresponding to the guide vane based on the historical guide vane parameters and the guide vane blade parameters.

[0014] Optionally, the determining module is configured to determine the parameter difference between each of the historical guide vane parameters and the guide vane blade parameters; cluster the historical guide vane parameters and the guide vane blade parameters according to the parameter difference to obtain multiple cluster sets; and determine the fault cause corresponding to the guide vane according to the multiple cluster sets.

[0015] Optionally, the determining module is further configured to determine the target probability of the guide vane experiencing the cause of failure in the future time based on the historical guide vane parameters, the guide vane blade parameters, and the cause of failure; The device further includes: The generation module is used to generate a corresponding guide vane maintenance scheme based on the target probability.

[0016] Thirdly, this disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the guide vane detection method provided in the first aspect of this disclosure.

[0017] Fourthly, this disclosure provides an electronic device, comprising: a memory storing a computer program thereon; and a processor for executing the computer program in the memory to implement the steps of the guide vane detection method provided in the first aspect of this disclosure.

[0018] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the guide vane detection method provided in the first aspect of this disclosure.

[0019] In the technical solution provided in this disclosure, firstly, the guide vane blade parameters at the target time are obtained. Secondly, the guide vane feature vector corresponding to the guide vane blade parameters is determined. Finally, the guide vane feature vector is input into a pre-constructed target detection model to obtain the operating status result of the guide vane output by the target detection model; the operating status result is used to characterize whether the guide vane is in a normal operating state or an abnormal operating state. The target detection model consists of an encoder and a generative adversarial network (GAN). The encoder generates corresponding encoded features based on the guide vane feature vector, and the GAN predicts the encoded features to generate the corresponding operating status result. Through the above method, the guide vane blade parameters at the target time can be obtained, and the corresponding guide vane feature vector can be generated based on the guide vane blade parameters. Then, the guide vane feature vector is input into the target detection model. Through the forward propagation and adversarial generation mechanism of the target detection model, the current operating state is analyzed in real time, thereby accurately identifying the operating status result of the guide vane, ensuring the real-time performance and accuracy of abnormal operation detection of the guide vane, greatly improving the accuracy and reliability of abnormal detection, and reducing false positives and false negatives.

[0020] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a method for detecting a guide vane according to an exemplary embodiment.

[0022] Figure 2 This is a flowchart illustrating another method for detecting guide vanes according to an exemplary embodiment.

[0023] Figure 3 This is a flowchart illustrating another method for detecting guide vanes according to an exemplary embodiment.

[0024] Figure 4 This is a flowchart illustrating a method for detecting a guide vane according to an exemplary embodiment.

[0025] Figure 5 This is a block diagram illustrating a guide vane detection device according to an exemplary embodiment.

[0026] Figure 6 This is a block diagram illustrating another guide vane detection device according to an exemplary embodiment.

[0027] Figure 7 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0028] The specific embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit this disclosure.

[0029] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily construed as referring to a specific order or sequence. Furthermore, in the description with reference to the accompanying drawings, the same reference numerals in different drawings denote the same elements.

[0030] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0031] In the description of this disclosure, unless otherwise stated, "multiple" means two or more, and other quantifiers are similar; "at least one," "one or more," or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one 'a' can represent any number of 'a's; as another example, one or more of a, b, and c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple; "and / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. The character " / " indicates that the preceding and following related objects are in an "or" relationship.

[0032] Although operations or steps are described in a specific order in the accompanying drawings in the embodiments of this disclosure, it should not be construed as requiring these operations or steps to be performed in the specific order or serial order shown, or requiring all of the shown operations or steps to be performed to obtain the desired result. In the embodiments of this disclosure, these operations or steps may be performed serially; they may be performed in parallel; or a portion of these operations or steps may be performed.

[0033] Before introducing the guide vane detection method, apparatus, storage medium, electronic device, and program product provided in this disclosure, the application scenarios involved in the various embodiments of this disclosure are first introduced. This disclosure can be applied to scenarios where guide vanes are in operation. During long-term operation, guide vanes are prone to failure due to mechanical wear, hydraulic system malfunctions, or environmental changes. Traditional guide vane fault detection methods usually rely on simple thresholds or rules, resulting in low detection accuracy and a tendency to miss or false alarms. In addition, many methods lack real-time monitoring and dynamic adjustment capabilities, making it impossible to identify and respond to abnormal situations during guide vane operation in a timely manner.

[0034] In related technologies, most guide vane fault detection methods are based on simple physical models and empirical rules, which have limited ability to identify faults under complex operating conditions. For example, threshold-based detection methods only trigger alarms when parameters exceed preset ranges, failing to consider the correlation between parameters and fault development trends, resulting in frequent missed or false alarms. Furthermore, while some methods introduce data-driven analysis, they lack effective data preprocessing and model optimization, leading to low detection accuracy and reliability.

[0035] Furthermore, in fault diagnosis, relevant technologies mainly rely on statistical analysis of historical fault data and expert experience, lacking a systematic fault cause analysis model. This method often fails to comprehensively and accurately identify the root cause of the fault and easily overlooks potential fault factors. In addition, traditional methods often cannot combine real-time monitoring data and historical data for comprehensive analysis, resulting in insufficient timeliness and accuracy in fault cause analysis.

[0036] To address the aforementioned technical problems, this invention provides a method, apparatus, storage medium, electronic device, and program product for detecting guide vanes. This method acquires the guide vane blade parameters at a target time and generates corresponding guide vane feature vectors based on these parameters. The guide vane feature vectors are then input into a target detection model. Through forward propagation and adversarial generation mechanisms within the target detection model, the current operating state is analyzed in real time, thereby accurately identifying the guide vane's operating state. This ensures the real-time performance and accuracy of abnormal guide vane operation detection, significantly improving the precision and reliability of anomaly detection and reducing false negatives and missed positives.

[0037] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0038] Figure 1 This is a flowchart illustrating a method for detecting a guide vane according to an exemplary embodiment, such as... Figure 1 As shown, the method may include the following steps: In step S101, the guide vane blade parameters at the target time are obtained.

[0039] The guide vane parameters may include, but are not limited to, the guide vane angle data θ(t), velocity data v(t), vibration data a(t), and temperature data T(t). In practical applications, the angle data θ(t) can be detected by setting an angle sensor on the guide vane, the velocity data can be detected by setting a velocity sensor, the vibration data can be detected by setting a vibration sensor, and the temperature data can be detected by setting a temperature sensor.

[0040] In some embodiments, an angle sensor can be mounted on the rotation shaft of the guide vane to detect the rotation angle of the guide vane and collect angle data θ(t):

[0041] Where θ(t) is the angle data, ω(t) is the angular velocity, and θ0 is the initial angle of the guide vane.

[0042] A velocity sensor can be installed on the moving part of the guide vane to detect the linear velocity of the guide vane and collect velocity data v(t):

[0043] Where v(t) is the velocity data, θ(t) is the angle data, and r is the radius of the guide vane.

[0044] Vibration sensors can be installed on key structural parts of the guide vane to detect the vibration amplitude and frequency of the guide vane and collect vibration data a(t):

[0045] Where a(t) represents the vibration data, a x (t), a y (t) and a z (t) represent the vibration components of the guide vane in the x, y, and z axes, which can be obtained, for example, based on the center position of the guide vane.

[0046] Temperature sensors can be installed on the surface or in key internal parts of the guide vane to detect the temperature of the guide vane and collect temperature data T(t).

[0047] After acquiring angle, velocity, vibration, and temperature data of the guide vane at multiple target times, the acquired data can be preliminarily processed to generate a guide vane operation dataset. For example, this guide vane operation dataset may include:

[0048] Among them, D initialThis is the guide vane operation dataset, where θ(t) represents angle data, v(t) represents velocity data, a(t) represents vibration data, T(t) represents temperature data, and t represents the time variable, t0, t1, ..., t n These are discrete time points (i.e., multiple target times).

[0049] In step S102, the guide vane feature vector corresponding to the guide vane blade parameter is determined.

[0050] In this step, the guide vane blade parameters can be pre-processed to obtain the target guide vane parameters. Then, the guide vane feature vector corresponding to the target guide vane parameters is determined. The pre-processing includes one or more of the following methods: denoising, normalization, and removal of outlier data. For example, the guide vane feature vector can be represented as X(t) = (θ(t), v(t), a(t), T(t)).

[0051] In some embodiments, when multiple processing methods are included, each processing method can be performed sequentially, that is, the later processing method is performed based on the previous processing method. For example, deleting abnormal data may include deleting outliers or missing values. Accordingly, the guide vane blade parameters can be cleaned to delete outliers or missing values. Further, the cleaned guide vane blade parameters can be denoised, for example, by using a moving average method to smooth the data to obtain denoised guide vane blade parameters. Then, the denoised guide vane blade parameters can be standardized (i.e., normalized) to normalize each data feature to the [0, 1] interval to obtain the target guide vane parameters.

[0052] In other embodiments, the aforementioned preset processing can also be performed on the data in the guide vane operation dataset. For example, taking the guide vane operation dataset as an example, the normalization formula may include:

[0053] Where, x norm x(t) represents the standardized data value, and x(t) represents the denoised data value. min and x max These are the minimum and maximum values ​​in the dataset, respectively.

[0054] After performing the same pre-processing on the aforementioned guide vane operation dataset, a pre-processed guide vane operation dataset is generated, which can be used for subsequent guide vane operation state model analysis. For example, the pre-processed guide vane operation dataset can be represented as:

[0055] In step S103, the guide vane feature vector is input into a pre-built target detection model to obtain the operating status result of the guide vane output by the target detection model.

[0056] The operational status result indicates whether the guide vane is in normal or abnormal operation. When the operational status result indicates that the guide vane is in an abnormal operation, it can be determined that a fault has occurred. At this time, based on the operational status result, corresponding alarm information can be generated to remind relevant technicians to promptly investigate and repair the guide vane's operational status. Simultaneously, the guide vane feature vector corresponding to the operational status result can be recorded and stored as a basis for subsequent fault analysis and maintenance.

[0057] In some embodiments, the target detection model consists of an encoder and a generative adversarial network (GAN). The encoder generates corresponding encoded features based on the guide vane feature vector, and the GAN predicts these encoded features to generate corresponding runtime results. The GAN includes a generator G. leaf Discriminator D leaf .

[0058] For example, this object detection model can be represented as M VAE-GAN Target detection model M VAE-GAN By using forward propagation and adversarial generative mechanisms, the current input data X(t) (i.e., the guide vane feature vector) can be analyzed to generate the corresponding running state result Y. pred (t), where Y pred (t)=1 indicates normal operating state, Y pred (t)=0 indicates an abnormal operating state. In one implementation, this Y pred (t) can be represented as:

[0059] Among them, Y pred (t) represents the running state result, σ is the activation function, and w is the weight vector. Let G be the kernel function, b be the bias term, and G be the kernel function. leaf (z) represents the generated data, and X(t) represents the guide vane feature vector.

[0060] In practical applications, the above method can be used to continuously monitor the operating status of the guide vanes. The guide vane feature vector at each time point t is input into the target detection model, thereby dynamically updating the operating status results and ensuring the real-time performance and accuracy of guide vane fault detection. At this point, the operating status result Y... pred (t+Δt) can be expressed as:

[0061] Where Δt is the time increment, D leaf G represents the discriminator. leaf (z) represents the generated data from the generator, and X(t+Δt) represents the guide vane feature vector.

[0062] Using the above method, the guide vane blade parameters at the target time can be obtained, and a corresponding guide vane feature vector can be generated based on these parameters. Then, the guide vane feature vector is input into the target detection model. Through the forward propagation and adversarial generation mechanism of the target detection model, the current operating state is analyzed in real time, thereby accurately identifying the guide vane's operating state. This ensures the real-time performance and accuracy of abnormal operation detection, greatly improving the precision and reliability of anomaly detection and reducing false negatives and missed positives.

[0063] The construction of the above target detection model will be explained in detail below.

[0064] In some embodiments, the target detection model can be a variational autoencoder-generative adversarial network (GAN) combination model. An improved variational autoencoder-GAN combination model is selected as the machine learning algorithm. By introducing a discriminator in the GAN, the ability to detect anomalies is enhanced, thereby improving the generation and discrimination capabilities of the target detection model.

[0065] First, the pre-processed guide vane running dataset D can be... normalized The model is divided into a training sample set and a test sample set. The training sample set is used to train the model in the early stage, and the test sample set is used to optimize and improve the model in the later stage.

[0066] Secondly, define the input feature vector of the object detection model (i.e., the sample guide vane feature vector):

[0067] Accordingly, an output label is set for each sample guide vane feature vector, Y(t) represents the operating status label of the guide vane, where Y(t)=1 indicates normal operating status and Y(t)=0 indicates abnormal operating status.

[0068] Then, a variational autoencoder is trained using the training sample set to learn the latent representation of the guide vane's normal operating state, thereby optimizing the model parameters:

[0069] The constraints include:

[0070] Among them, D KL Let q(z|X) be the Kullback-Leibler divergence, q(z|X) be the approximate posterior distribution, p(z) be the prior distribution, p(X|z) be the generative model, and X be the value of the Kullback-Leibler divergence.i Let μ(X) be the feature vector of the sample guide vane. i |z) and σ 2 (X i |z) represents the mean and variance of the generated values, respectively.

[0071] Furthermore, the generative adversarial network may include a generator G. leaf Discriminator D leaf Generator G leaf This is used to generate the operating data of the guide vane. After inputting the noise variable z, the operating status data of the guide vane is generated. The noise variable z is sampled from the prior distribution p(z). Discriminator D leaf Used to distinguish between real guide vane operation data X and guide vane operation status data generated by the generator. Output the discrimination probability D leaf (X) represents the probability that the input data is the real data, which is also the output running status result.

[0072] In one possible implementation, the generative adversarial network can be pre-trained in the following way: Step A: Obtain a preset training sample set, which includes multiple sample guide vane feature vectors and the running status label corresponding to each sample guide vane feature vector.

[0073] The operation status label includes either normal operation status or abnormal operation status.

[0074] Step B involves training a preset training model based on the multiple sample guide vane feature vectors and the running state label corresponding to each sample guide vane feature vector, and then using the trained preset training model as the generative adversarial network.

[0075] Specifically, the pre-trained model may include a generator and a discriminator, and adversarial training is performed on the normal operating state and abnormal operating state generated by the pre-trained model in the latent space, wherein the optimization objective function is:

[0076] Where X is the feature vector of the sample guide vane, p data (X) represents the true data distribution (i.e., the running status label), Z represents the noise variable, p(z) represents the prior distribution, and G represents the noise variable. leaf (z) represents the generated data, D leaf (X) represents the probability that the input data is the real data.

[0077] In this way, an object detection model M can be established using a combination model of a variational autoencoder and a generative adversarial network trained against adversarial forces. VAE-GAN Target detection model M VAE-GANThrough adversarial generation and discrimination mechanisms, the normal and abnormal operating states of the guide vanes can be identified and detected in real time.

[0078] In some embodiments, for generator G leaf The optimization includes: introducing a conditional generation model, using historical operating data of the guide vane as generation conditions to generate future operating data for the guide vane, and using an improved first loss function combined with mean squared error and structural similarity metrics to optimize the quality of the generated data. The first loss function can be expressed as:

[0079] in, λ1 and λ2 are the first loss function, SSIM is the structural similarity measure function, X is the sample guide vane feature vector, and G(z) is the data generated by the generator.

[0080] For discriminant D leaf The optimizations include: introducing a multi-scale discriminator to evaluate the authenticity of generated data at different scales, enhancing the ability to discriminate details, using an improved adversarial loss function, and combining it with a contrastive learning method to improve the robustness of the discriminator. The adversarial loss function can be expressed as:

[0081] in, The loss function is adversarial, and λ3 is the weight coefficient for contrastive learning. For the negative samples generated in the comparative learning, G(z) is the data generated by the generator, X is the sample guide vane feature vector, and D() represents the discriminator.

[0082] In practical applications, as the number of guide vane blade parameters collected during use accumulates, a corresponding training sample set can be periodically established based on the latest collected guide vane blade parameters, and the target detection model can be regularly updated and optimized, enabling the target detection model to adapt to different operating conditions and environmental changes.

[0083] In current guide vane technology, determining the cause of guide vane failure typically relies on statistical analysis of historical failure data and the experience of experts. This approach suffers from poor timeliness in cause analysis and fails to comprehensively and accurately identify the root cause of the failure, easily overlooking potential contributing factors. To address these technical problems, in some embodiments, such as... Figure 2 As shown, the method may further include the following steps: In step S104, if the operating status result includes an abnormal operating status, the historical guide vane parameters within the historical time period prior to the target time are obtained.

[0084] The historical guide vane parameters can be expressed as:

[0085] The guide vane blade parameters at the target time can be expressed as:

[0086] In step S105, the cause of the fault corresponding to the guide vane is determined based on the historical guide vane parameters and the guide vane blade parameters.

[0087] In this step, firstly, the parameter difference between each historical guide vane parameter and the guide vane blade parameter can be determined.

[0088] For example, a feature comparison analysis can be performed on the historical guide vane parameters and the guide vane blade parameters at the target time to calculate the difference vector ΔX(t) for each feature:

[0089] Where ΔX(t) is the feature difference vector at time t.

[0090] Secondly, based on the parameter difference, the historical guide vane parameters and the guide vane blade parameters are clustered to obtain multiple cluster sets.

[0091] For example, the parameter differences, i.e. the feature difference vector ΔX(t), can be clustered to obtain multiple cluster sets.

[0092] Then, based on these multiple clusters, the cause of the guide vane failure was determined.

[0093] In this step, statistical analysis methods can be used to perform cluster analysis on the feature difference vector ΔX(t), identify anomalous feature clusters, and determine the corresponding fault types. For example, clusters with fewer than a preset threshold can be considered anomalous feature clusters. Alternatively, anomalous feature clusters can be determined based on the distance between each cluster and its cluster center, or other methods can be used, depending on actual needs. Fault types can be assigned based on pre-experimental analysis in the laboratory, assigning corresponding fault types to clusters with different distances from their cluster centers. Furthermore, the identified anomalous feature clusters can be used to determine the corresponding fault types.

[0094] In some embodiments, clustering based on anomaly features can be used to analyze the causes of failures using Bayesian networks, constructing a probability graph of failure causes G=(V, E), where V is a set of nodes representing different failure causes, and E is a set of edges representing the relationships between failure causes.

[0095] Where P(F|ΔX) is the probability of the cause of failure given a feature difference vector, and F i The cause of the fault is Pa(F) i ) is F i The set of parent nodes.

[0096] After obtaining the fault cause probability map, a corresponding fault cause analysis report can be generated, which includes the fault type, fault occurrence time, parameter differences, possible fault causes, and corresponding fault occurrence probabilities. This fault cause analysis report is then stored in a database and sent to relevant maintenance personnel for fault handling and maintenance decision-making reference.

[0097] In this way, this disclosure can not only trigger a fault alarm mechanism, but also perform a preliminary diagnosis of the fault cause. By analyzing historical guide vane parameters and guide vane blade parameters, it provides a detailed fault cause analysis report, helping maintenance personnel to quickly locate problems, providing accurate fault causes and detailed analysis reports, thereby improving the efficiency and accuracy of fault handling.

[0098] In addition, after obtaining the fault cause analysis report, this disclosure also provides a method for generating the corresponding guide vane maintenance plan, so as to provide relevant maintenance personnel with a more detailed and comprehensive analysis report. Specifically, such as Figure 3 As shown, the method may further include the following steps: In step S106, based on the historical guide vane parameters, the guide vane blade parameters, and the cause of the failure, the target probability of the guide vane experiencing the cause of the failure in the future is determined.

[0099] In this step, firstly, based on the fault causes and their corresponding probabilities in the fault cause analysis report, a Bayesian time series model can be used for modeling. Combining the fault cause probabilities mentioned above with the current data trend, a time series model can be constructed:

[0100] in, For a time series model, P(X) t+1 |X t F i ) indicates that in fault F i Probability of future trends under certain conditions Let ΔX(t) be the set of all possible faults, and let ΔX(t) be the feature difference vector at time t.

[0101] Then, a multinomial regression model can be used to predict the future trend of guide vane parameters. By combining historical guide vane parameters and weighted updates of guide vane blade parameters, Bayesian estimates of the regression coefficients can be obtained.

[0102] Wherein, P(w|X historical X current Let ) be the likelihood function of the data, w be the prior parameter, and P(w) be the prior distribution of the regression coefficients.

[0103] Finally, by combining the predicted trends and the failure cause analysis report, Markov random fields can be used to calculate the joint probability of future failures, i.e., the target probability:

[0104] in, Let Z be the target probability, Z be the normalization constant, and θ be the θ value. i and θ i，j f is a parameter i and f i，j Let F be the characteristic function and F be the cause of the fault. This is the operational data for the guide vane status.

[0105] In step S107, a corresponding guide vane maintenance scheme is generated based on the target probability.

[0106] In this step, based on the target probability of future failures, a multi-objective optimization algorithm can be used to generate preventative maintenance recommendations, defining the optimization objective functions for maintenance time, content, and priority:

[0107] in, Let C(t) be the target probability, C(t) be the failure cost function, and E(Time) be the failure cost function. t P(Priority) represents the expected maintenance time. t To maintain the probability distribution of priorities, λ4 and λ5 are weighting coefficients.

[0108] In other words, when the above optimization objective function reaches its minimum value, a corresponding preset maintenance suggestion, i.e., a guide vane maintenance scheme, can be generated based on the expected value of the corresponding maintenance time and the probability distribution of the maintenance priority.

[0109] Then, based on the generated preventative maintenance recommendations, a detailed maintenance plan can be developed using dynamic programming algorithms, including maintenance steps and resource allocation:

[0110] Among them, R(Steps) t Resources t To maintain the revenue function, C(Steps) t Resources t Let be the maintenance cost function. By optimizing the solution, the optimal maintenance plan can be obtained.

[0111] The guide vane detection method provided in this disclosure overcomes the shortcomings of related technologies. Through real-time monitoring and intelligent analysis of guide vane blade parameters, it achieves high-precision fault detection and diagnosis. It can not only promptly identify abnormal conditions during guide vane operation but also provide detailed fault cause analysis and preventative maintenance recommendations, thereby improving the reliability and safety of the equipment.

[0112] Figure 4 This is a flowchart illustrating a method for detecting a guide vane according to an exemplary embodiment, such as... Figure 4 As shown, in the daily operation of a hydropower station, the guide vane, as an important component of the turbine, directly affects the efficiency and safety of the entire equipment. However, due to long-term operation, factors such as guide vane wear, hydraulic system failures, and environmental changes frequently cause abnormal conditions in the guide vane. To improve the accuracy and real-time performance of guide vane fault detection and implement effective preventive maintenance strategies, the following steps can be adopted: S1. Data Acquisition: Angle sensors, speed sensors, vibration sensors, and temperature sensors can be installed on the guide vane to collect guide vane blade parameters in real time and construct a guide vane operation dataset. S2. Data Preprocessing: After denoising, normalizing, and deleting outlier data from the guide vane operation dataset, a training sample set and a test sample set are obtained. S3. Model Training: Based on the training sample set and the test sample set, a pre-set training model is trained to obtain the corresponding target detection model. S4. Real-time Monitoring: In practical application scenarios, the guide vane blade parameters can be acquired in real time, converted into corresponding guide vane feature vectors, and input into the pre-constructed target detection model. S5. Fault Judgment: Obtain the operating status result of the guide vane output by the target detection model. This operating status result is used to characterize whether the guide vane is in a normal operating state or an abnormal operating state. When the operating status result indicates that the guide vane is in an abnormal operating state, it can be determined that the current guide vane has failed. S6. Data Storage: If the model determines that the current state is abnormal, it immediately triggers a fault alarm mechanism, records and stores the current abnormal data. S7. Fault Diagnosis: The system combines historical guide vane parameters and current guide vane blade parameters to perform a preliminary diagnosis of the fault cause and provides a detailed fault cause analysis report. S8. Maintenance Recommendations: Based on the fault cause analysis report and the trend of operating data changes (i.e., target probability), preventive maintenance recommendations are generated, and corresponding guide vane maintenance plans are formulated. S9. Model Update: The target detection model is regularly updated and optimized so that it can adapt to different operating conditions and environmental changes.

[0113] The following example illustrates the guide vane detection method using a specific application scenario. One day, during routine operation, the vibration parameters of the guide vane were found to be significantly outside the normal range, and the temperature parameter showed an abnormal increase. Real-time data vectors were input into the target detection model to analyze the current operating status of the guide vane. The predicted values ​​showed anomalies, triggering a fault alarm mechanism. Subsequently, a preliminary diagnosis of the fault was initiated. By analyzing historical guide vane parameters and current guide vane blade parameters, the parameter differences were calculated, and statistical analysis methods were used to perform cluster analysis on the parameter differences, identifying abnormal clusters of vibration and temperature parameters. Combined with a Bayesian network, the system generated a fault cause probability map, determining that the hydraulic system was highly likely to be faulty. Based on this analysis, the system generated a detailed fault cause analysis report, indicating that the hydraulic system might have oil leaks or blockages, and recommending immediate inspection and maintenance of the hydraulic system. Subsequently, based on the fault cause analysis report and operating data trends, the system used a Bayesian time series model and a multi-objective optimization algorithm to generate preventative maintenance recommendations, including specific inspection times, maintenance content, and priorities, and formulated a detailed maintenance plan.

[0114] To verify the effectiveness of the method provided in this disclosure, laboratory verification yielded the comparative data shown in Table 1 below. It can be seen that the guide vane detection method provided in this disclosure significantly outperforms methods in related technologies in terms of fault detection accuracy, fault response time, maintenance plan rationality, and equipment lifespan. Furthermore, the false alarm rate and missed alarm rate are significantly reduced, fully demonstrating the progressiveness and practicality of this disclosure in guide vane fault detection and diagnosis. By applying the method provided in this disclosure, hydropower stations not only improve equipment reliability and safety but also significantly reduce maintenance costs and extend equipment lifespan, achieving good economic and social benefits.

[0115]

[0116] Table 1 Figure 5 This is a block diagram illustrating a guide vane detection device according to an exemplary embodiment, such as... Figure 5 As shown, the device 200 includes: The acquisition module 201 is used to acquire the guide vane blade parameters at the target time; The determining module 202 is used to determine the guide vane feature vector corresponding to the guide vane blade parameters; The detection module 203 is used to input the feature vector of the guide vane into a pre-built target detection model to obtain the operating status result of the guide vane output by the target detection model; the operating status result is used to characterize whether the guide vane is in normal operating state or abnormal operating state. The target detection model consists of an encoder and a generative adversarial network. The encoder is used to generate corresponding encoded features based on the guide vane feature vector, and the generative adversarial network is used to predict the encoded features and generate the corresponding running status results.

[0117] Optionally, the determining module 202 is used to perform preset processing on the guide vane blade parameters to obtain target guide vane parameters; the preset processing includes one or more of the following processing methods: noise reduction processing, normalization processing and deletion of abnormal data; and to determine the guide vane feature vector corresponding to the target guide vane parameters.

[0118] Optionally, the generative adversarial network is pre-trained in the following manner: Obtain a preset training sample set, which includes multiple sample guide vane feature vectors and a running status label corresponding to each sample guide vane feature vector; the running status label includes normal running status or abnormal running status. The preset training model is trained based on the multiple sample guide vane feature vectors and the running state label corresponding to each sample guide vane feature vector, and the trained preset training model is used as the generative adversarial network.

[0119] Optionally, the acquisition module 201 is further configured to acquire historical guide vane parameters within a historical time period prior to the target time if the operating status result includes an abnormal operating status. The determining module 202 is also used to determine the cause of the fault corresponding to the guide vane based on the historical guide vane parameters and the guide vane blade parameters.

[0120] Optionally, the determining module 202 is used to determine the parameter difference between each historical guide vane parameter and the guide vane blade parameter; cluster the historical guide vane parameter and the guide vane blade parameter according to the parameter difference to obtain multiple cluster sets; and determine the fault cause corresponding to the guide vane according to the multiple cluster sets.

[0121] Optionally, the determining module 202 is further configured to determine the target probability of the guide vane experiencing the cause of failure in the future time based on the historical guide vane parameters, the guide vane blade parameters, and the cause of failure; like Figure 6 As shown, the device 200 also includes: The generation module 204 is used to generate a corresponding guide vane maintenance scheme based on the target probability.

[0122] Using the aforementioned device, the guide vane blade parameters at the target time can be acquired, and a corresponding guide vane feature vector can be generated based on these parameters. Subsequently, the guide vane feature vector is input into the target detection model. Through the forward propagation and adversarial generation mechanism of the target detection model, the current operating state is analyzed in real time, thereby accurately identifying the guide vane's operating state. This ensures the real-time performance and accuracy of abnormal operation detection, significantly improving the precision and reliability of anomaly detection and reducing false negatives and missed positives.

[0123] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0124] Figure 7 This is a block diagram illustrating an electronic device 300 according to an exemplary embodiment. Figure 7 As shown, the electronic device 300 may include a processor 301 and a memory 302. The electronic device 300 may also include one or more of a multimedia component 303, an input / output (I / O) interface 304, and a communication component 305.

[0125] The processor 301 controls the overall operation of the electronic device 300 to complete all or part of the steps in the guide vane detection method described above. The memory 302 stores various types of data to support the operation of the electronic device 300. This data may include, for example, instructions for any application or method operating on the electronic device 300, and application-related data such as contact data, sent and received messages, pictures, audio, video, etc. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 303 may include a screen and audio components. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in memory 302 or transmitted via communication component 305. The audio component also includes at least one speaker for outputting audio signals. I / O interface 304 provides an interface between processor 301 and other interface modules, such as a keyboard, mouse, buttons, etc. These buttons may be virtual or physical buttons. Communication component 305 is used for wired or wireless communication between the electronic device 300 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, NB-IoT, eMTC, or other 5G technologies, or combinations thereof, is not limited here. Therefore, the corresponding communication component 305 may include: a Wi-Fi module, a Bluetooth module, an NFC module, etc.

[0126] In an exemplary embodiment, the electronic device 300 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the guide vane detection method described above.

[0127] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the guide vane detection method described above. For example, the computer-readable storage medium may be the memory 302 including the program instructions described above, which may be executed by the processor 301 of the electronic device 300 to complete the guide vane detection method described above.

[0128] In another exemplary embodiment, a computer program product is also provided, comprising a computer program executable by a programmable device, the computer program having a code portion for performing the above-described guide vane detection method when executed by the programmable device.

[0129] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0130] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0131] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

Claims

1. A method for detecting guide vanes, characterized in that, The method includes: Obtain the guide vane blade parameters at the target time; Determine the guide vane feature vector corresponding to the guide vane blade parameters; The guide vane feature vector is input into a pre-constructed target detection model to obtain the operating status result of the guide vane output by the target detection model; the operating status result is used to characterize whether the guide vane is in a normal operating state or an abnormal operating state. The target detection model consists of an encoder and a generative adversarial network. The encoder is used to generate corresponding encoded features based on the guide vane feature vector, and the generative adversarial network is used to predict the encoded features and generate corresponding running status results.

2. The method according to claim 1, characterized in that, The step of determining the guide vane feature vector corresponding to the guide vane blade parameters includes: The guide vane blade parameters are pre-processed to obtain target guide vane parameters; the pre-processing includes one or more of the following processing methods: noise reduction processing, normalization processing, and deletion of abnormal data; Determine the guide vane feature vector corresponding to the target guide vane parameters.

3. The method according to claim 1, characterized in that, The generative adversarial network is pre-trained in the following manner: Obtain a preset training sample set, which includes multiple sample guide vane feature vectors and a running status label corresponding to each sample guide vane feature vector; the running status label includes normal running status or abnormal running status. The preset training model is trained based on the multiple sample guide vane feature vectors and the running state label corresponding to each sample guide vane feature vector, and the trained preset training model is used as the generative adversarial network.

4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: If the operating status result includes an abnormal operating status, obtain the historical guide vane parameters within the historical time period prior to the target time. Based on the historical guide vane parameters and the guide vane blade parameters, determine the cause of the guide vane failure.

5. The method according to claim 4, characterized in that, The step of determining the cause of the guide vane failure based on the historical guide vane parameters and the guide vane blade parameters includes: Determine the parameter difference between each of the historical guide vane parameters and the guide vane blade parameters; Based on the parameter differences, the historical guide vane parameters and the guide vane blade parameters are clustered to obtain multiple cluster sets; Based on the multiple cluster sets, the cause of the failure corresponding to the guide vane is determined.

6. The method according to claim 5, characterized in that, The method further includes: Based on the historical guide vane parameters, the guide vane blade parameters, and the cause of failure, determine the target probability of the guide vane experiencing the cause of failure in the future time. Based on the target probability, a corresponding guide vane maintenance scheme is generated.

7. A detection device for guide vanes, characterized in that, The device includes: The acquisition module is used to acquire the guide vane blade parameters at the target time. The determination module is used to determine the guide vane feature vector corresponding to the guide vane blade parameters; The detection module is used to input the feature vector of the guide vane into a pre-constructed target detection model to obtain the operating status result of the guide vane output by the target detection model; the operating status result is used to characterize whether the guide vane is in a normal operating state or an abnormal operating state. The target detection model consists of an encoder and a generative adversarial network. The encoder is used to generate corresponding encoded features based on the guide vane feature vector, and the generative adversarial network is used to predict the encoded features and generate corresponding running status results.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images