A water turbine guide vane spraying control method and system based on a neural network model

Abstract

The application provides a water turbine guide vane spraying control method and system based on a neural network model, and relates to the technical field of water turbine guide vane coating spraying. The method comprises the following steps: acquiring historical operation state data of a water turbine guide vane during operation; determining preliminary spraying thickness information based on the historical operation state data of the water turbine guide vane during operation; performing secondary spraying on a test water turbine guide vane after preliminary spraying based on secondary test spraying information of each wear area, and acquiring operation state data of the test water turbine guide vane during operation after secondary spraying; determining a target spraying scheme based on the operation state data of the test water turbine guide vane during operation after secondary spraying; and performing spraying control of a graphene ceramic gold coating on water turbine guide vanes of a product line based on the target spraying scheme. The method can accurately formulate an adaptive graphene ceramic gold coating spraying scheme according to the actual wear state of the water turbine guide vane.

Description

Technical Field

[0001] This invention relates to the field of turbine guide vane coating technology, specifically to a turbine guide vane coating control method and system based on a neural network model. Background Technology

[0002] The guide vanes of a hydro-turbine generator unit are key control components for regulating flow and output. They are subjected to the combined effects of high-speed, sand-laden water flow, cavitation erosion, and electrochemical corrosion, resulting in significant non-uniform wear distribution on their surface. This severely impacts the sealing performance, control accuracy, and operational stability of the unit. To address this issue, surface strengthening with high-performance coatings has become an important technical approach. Among these, graphene-ceramic composite coatings, with their excellent mechanical strength, self-lubricating properties, and corrosion resistance, show significant potential in improving the wear resistance and cavitation erosion resistance of guide vanes. However, existing spraying processes generally use uniform coating parameters, which cannot adapt to the differentiated protection requirements arising from the uneven hydrodynamic distribution on the guide vane surface. This leads to insufficient local protection or material redundancy, making it difficult to fully utilize the performance and economic advantages of graphene coatings. In particular, it is crucial to identify and classify the surface based on actual wear data from guide vane operation, and dynamically optimize key spraying parameters such as coating thickness and graphene content in different areas to achieve the optimal balance between protective effectiveness and cost control. The lack of systematic intelligent decision-making and control methods currently hinders the large-scale engineering application of this advanced coating technology in the long-term protection of guide vanes.

[0003] Therefore, how to accurately formulate a suitable graphene ceramic coating spraying scheme based on the actual wear condition of the turbine guide vanes is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to accurately formulate a suitable graphene ceramic coating spraying scheme based on the actual wear state of the turbine guide vanes.

[0005] According to a first aspect, the present invention provides a method for controlling the spraying of a turbine guide vane based on a neural network model, comprising: acquiring historical operating status data of the turbine guide vane during operation; determining preliminary coating thickness information based on the historical operating status data of the turbine guide vane during operation; controlling a spraying robot to perform preliminary spraying of a graphene ceramic coating on a test turbine guide vane based on the preliminary coating thickness information, and acquiring operating status data of the test turbine guide vane after preliminary spraying during operation; determining multiple wear area information of the test turbine guide vane after preliminary spraying based on the operating status data of the test turbine guide vane after preliminary spraying; and further... The secondary test coating information for each wear area of ​​the test turbine guide vane after the initial coating is generated from multiple wear area information. The secondary test coating information for each wear area includes graphene content and coating thickness. Based on the secondary test coating information for each wear area, the test turbine guide vane after the initial coating is subjected to secondary coating, and the operating status data of the test turbine guide vane after secondary coating is obtained. Based on the operating status data of the test turbine guide vane after secondary coating, a target coating scheme is determined. Based on the target coating scheme, the coating control of graphene ceramic gold coating is performed on the turbine guide vanes of the product line.

[0006] In one possible implementation, determining multiple wear area information of the test turbine guide vanes after preliminary coating based on the operating status data of the test turbine guide vanes during operation includes: determining multiple wear point information of the test turbine guide vanes after preliminary coating based on the operating status data of the test turbine guide vanes during operation; clustering multiple wear point information of the test turbine guide vanes after preliminary coating to obtain multiple clusters; and determining multiple wear area information of the test turbine guide vanes after preliminary coating based on the multiple clusters.

[0007] In one possible implementation, determining the target coating scheme based on the operating status data of the test turbine guide vanes after the secondary coating includes: determining the wear information of each wear area after the secondary test coating based on the operating status data of the test turbine guide vanes after the secondary coating; constructing a secondary coating map, which includes multiple wear areas and multiple edges between the multiple wear areas, and the node features of each wear area are the wear information after the secondary test coating and the secondary test coating information of each wear area; processing the secondary coating map based on a graph neural network to obtain the secondary target coating information of each wear area; and determining the target coating scheme based on the preliminary coating thickness information and the secondary target coating information of each wear area.

[0008] In one possible implementation, generating secondary test coating information for each wear area based on the multiple wear area information of the test turbine guide vanes after the initial coating includes: determining multiple representative wear areas, secondary test coating information for each representative wear area, and coating scheme difference degree based on the multiple wear area information of the test turbine guide vanes after the initial coating; and generating secondary test coating information for other wear areas based on the secondary test coating information of each representative wear area and the coating scheme difference degree.

[0009] According to a second aspect, the present invention provides a turbine guide vane spraying control system based on a neural network model, comprising: an acquisition module for acquiring historical operating status data of turbine guide vanes during operation; a preliminary coating thickness determination module for determining preliminary coating thickness information based on the historical operating status data of turbine guide vanes during operation using an operation analysis model, wherein the operation analysis model is a deep neural network model; a preliminary coating control module for controlling a spraying robot to perform preliminary coating of graphene ceramic gold coating on test turbine guide vanes based on the preliminary coating thickness information, and acquiring operating status data of the test turbine guide vanes after preliminary coating during operation; and a wear area determination module for determining the wear area of ​​the test turbine guide vanes after preliminary coating based on the operating status data of the test turbine guide vanes after preliminary coating during operation. The system includes: a secondary test coating information generation module for generating secondary test coating information for each wear area based on the multiple wear area information of the test turbine guide vanes after the initial coating; a secondary coating control module for performing secondary coating on the test turbine guide vanes after the initial coating based on the secondary test coating information of each wear area, and acquiring the operating status data of the test turbine guide vanes after secondary coating; a target coating scheme determination module for determining the target coating scheme based on the operating status data of the test turbine guide vanes after secondary coating; and a coating control module for controlling the coating of graphene ceramic gold coating on the turbine guide vanes of the product line based on the target coating scheme.

[0010] In one possible implementation, the wear area determination module is further configured to: determine multiple wear point information of the test turbine guide vane after preliminary coating based on the operating status data of the test turbine guide vane during operation; cluster the multiple wear point information of the test turbine guide vane after preliminary coating to obtain multiple clusters; and determine multiple wear area information of the test turbine guide vane after preliminary coating based on the multiple clusters.

[0011] In one possible implementation, the target coating scheme determination module is further configured to: determine the wear information of each wear area after secondary test coating based on the operating status data of the test turbine guide vanes after secondary coating; construct a secondary coating map, the secondary coating map including multiple wear areas and multiple edges between the multiple wear areas, the node features of each wear area being the wear information after secondary test coating of each wear area and the secondary test coating information of each wear area; process the secondary coating map based on a graph neural network to obtain the secondary target coating information of each wear area; and determine the target coating scheme based on the preliminary coating thickness information and the secondary target coating information of each wear area.

[0012] In one possible implementation, the secondary test spraying information generation module is further configured to: determine multiple representative wear areas, secondary test spraying information for each representative wear area, and spraying scheme difference degree based on multiple wear area information of the test turbine guide vanes after the initial spraying; and generate secondary test spraying information for other wear areas based on the secondary test spraying information of each representative wear area and the spraying scheme difference degree.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring historical operating status data of turbine guide vanes during operation; determining preliminary coating thickness information using an operational analysis model based on the historical operating status data of turbine guide vanes during operation, the operational analysis model being a deep neural network model; controlling a coating robot to perform preliminary coating of graphene ceramic gold coating on test turbine guide vanes based on the preliminary coating thickness information, and acquiring operating status data of the test turbine guide vanes during operation after preliminary coating; and acquiring operating status data of the test turbine guide vanes after preliminary coating based on the preliminary coating thickness information. The operating status data of the turbine guide vanes during operation determines the information of multiple wear areas of the test turbine guide vanes after preliminary coating; based on the multiple wear area information of the test turbine guide vanes after preliminary coating, secondary test coating information is generated for each wear area, including graphene content and coating thickness; based on the secondary test coating information of each wear area, a secondary coating is performed on the test turbine guide vanes after preliminary coating, and the operating status data of the test turbine guide vanes after secondary coating is obtained; based on the operating status data of the test turbine guide vanes after secondary coating, a target coating scheme is determined; based on the target coating scheme, the coating control of graphene ceramic gold coating is performed on the turbine guide vanes of the product line.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned method for controlling the spraying of turbine guide vanes based on a neural network model. The method includes: acquiring historical operating status data of turbine guide vanes during operation; determining preliminary coating thickness information using an operational analysis model based on the historical operating status data of turbine guide vanes during operation, wherein the operational analysis model is a deep neural network model; controlling a spraying robot to perform preliminary spraying of a graphene ceramic coating on test turbine guide vanes based on the preliminary coating thickness information, and acquiring operating status data of the test turbine guide vanes after preliminary spraying; and acquiring operating status data of the test turbine guide vanes after preliminary spraying. The system uses operational status data to determine multiple wear areas of the test turbine guide vanes after initial coating. Based on this information, secondary coating information is generated for each wear area, including graphene content and coating thickness. A secondary coating is then applied to the test turbine guide vanes after initial coating, and operational status data is obtained. A target coating scheme is determined based on this operational status data. Finally, the graphene ceramic coating is controlled for the turbine guide vanes in the product line.

[0015] This invention provides a method and system for controlling the spraying of turbine guide vanes based on a neural network model. The method includes: acquiring historical operational status data of turbine guide vanes during operation; determining preliminary coating thickness information using an operational analysis model (a deep neural network model) based on the historical operational status data; controlling a spraying robot to perform preliminary graphene ceramic coating spraying on test turbine guide vanes based on the preliminary coating thickness information, and acquiring operational status data of the test turbine guide vanes after preliminary coating; determining multiple wear areas of the test turbine guide vanes after preliminary coating based on the operational status data of the test turbine guide vanes after preliminary coating; and determining the preliminary coating thickness based on the preliminary coating thickness information. Multiple wear areas of the pre-coated test turbine guide vanes are used to generate secondary test coating information for each wear area. This secondary test coating information for each wear area includes graphene content and coating thickness. Based on this secondary test coating information for each wear area, a secondary coating is applied to the pre-coated test turbine guide vanes, and operational status data of the pre-coated test turbine guide vanes during operation is obtained. A target coating scheme is determined based on the operational status data of the pre-coated test turbine guide vanes. Based on the target coating scheme, the graphene ceramic coating is controlled for the turbine guide vanes of the product line. This method can accurately formulate a suitable graphene ceramic coating coating scheme according to the actual wear state of the turbine guide vanes. Attached Figure Description

[0016] Figure 1 A schematic flowchart of a water turbine guide vane spraying control method based on a neural network model provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a water turbine guide vane provided in an embodiment of the present invention;

[0018] Figure 3 A flowchart illustrating a process for determining multiple wear areas of a test turbine guide vane after preliminary coating, as provided in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of a process for generating secondary test coating information for each wear area, provided by an embodiment of the present invention.

[0020] Figure 5 This is a schematic flowchart illustrating the process of determining a target spraying scheme according to an embodiment of the present invention.

[0021] Figure 6 This is a schematic diagram of a water turbine guide vane spraying control system based on a neural network model, provided as an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0023] In this embodiment of the invention, the following are provided: Figure 1 The method for controlling the spraying of water turbine guide vanes based on a neural network model is shown. The method includes steps S1 to S8:

[0024] Step S1: Obtain historical operating status data of the turbine guide vanes.

[0025] Turbine guide vanes are the core component of the guide vane mechanism in reaction turbines (such as axial-flow and mixed-flow turbines). Turbine guide vanes can have an airfoil cross-section and are evenly distributed around the runner. By adjusting the opening and closing angle of the turbine guide vanes, the flow channel can be formed and altered to control the flow rate, velocity, and direction of the water entering the runner, thereby achieving precise regulation of the turbine's output and rotational speed. Figure 2 This is a schematic diagram of a water turbine guide vane provided in an embodiment of the present invention.

[0026] Historical operating status data of turbine guide vanes is a collection of data organized in a time series, continuously recorded by sensors and monitoring systems over a period of time during which the turbine guide vanes operated.

[0027] Historical operating status data of the turbine guide vanes include records of water flow velocity at different time points, water pressure changes, guide vane opening and closing angles, guide vane surface stress and strain, unit vibration frequency and amplitude, and changes in sediment content in the working environment.

[0028] Step S2: Based on the historical operating status data of the turbine guide vanes, the preliminary coating thickness information is determined using an operating analysis model, which is a deep neural network model.

[0029] In some embodiments, an operational analysis model can be used to determine the initial coating thickness information. The operational analysis model is a deep neural network model. The input to the operational analysis model is the historical operating status data of the turbine guide vanes, and the output of the operational analysis model is the initial coating thickness information.

[0030] Deep neural network models include deep neural networks (DNNs). A deep neural network is a computational model that mimics the structure and function of biological neural networks. A deep neural network consists of an input layer, multiple hidden layers, and an output layer. In the hidden layers, neurons receive input from the previous layer, which is then weighted, summed, processed by an activation function, and passed to the next layer. Deep neural networks are capable of learning nonlinear relationships and feature representations in data.

[0031] The initial coating thickness information is the distribution information of the initial coating thickness of the graphene ceramic gold coating at different specific locations of the test turbine guide vanes, determined by the operational analysis model.

[0032] The preliminary coating thickness information clarifies the specific thickness value corresponding to each sprayable area in the guide vane (such as the guide vane edge, the middle stress area, the root non-stress area, the inlet and outlet sides, etc.). The thickness values ​​at different locations show a differentiated distribution based on the differences in the historical wear degree, stress magnitude, and water flow impact intensity at that location.

[0033] Preliminary coating thickness information is used to indicate the thickness parameters of the first basic coating operation.

[0034] Historical operating data of turbine guide vanes covers wear records under various conditions, including load, flow velocity, and operating duration. This data comprehensively presents the correlation between different operating conditions and the actual wear levels of the guide vanes. Coating thickness is a core factor affecting the protective effect of the coating, and there is a clear intrinsic correlation between the historical operating data of turbine guide vanes and the protective effect of the graphene ceramic coating. The differences in wear levels of the guide vanes under different operating conditions directly determine the required coating thickness for the corresponding areas. This historical data, containing the correlation between operating conditions and wear, provides rich learning samples for the operational analysis model.

[0035] Deep neural networks learn features layer by layer from historical turbine guide vane operating data, enabling them to uncover potential correlations between operating parameters and guide vane wear. By integrating operating data, the model can accurately capture the guide vane's coating thickness requirements under different operating conditions, and then output reasonable preliminary coating thickness information based on these learned patterns.

[0036] Step S3: Based on the preliminary coating thickness information, control the spraying robot to perform preliminary coating of graphene ceramic gold coating on the test turbine guide vane, and obtain the operating status data of the test turbine guide vane during operation after preliminary coating.

[0037] The operational status data of the test turbine guide vanes after preliminary coating are records of physical parameters collected in the actual operating environment after the coating is applied according to the preliminary coating thickness information, used to evaluate the coating performance.

[0038] The operational status data of the test turbine guide vanes after initial coating include hydraulic efficiency maintenance index data during operation, stress distribution on the guide vane surface, vibration spectrum, and temperature field data of the coating area.

[0039] Step S4: Based on the operating status data of the test turbine guide vanes after preliminary coating, determine the information of multiple wear areas of the test turbine guide vanes after preliminary coating.

[0040] In some embodiments, Figure 3 This is a flowchart illustrating a method for determining multiple wear areas of a test turbine guide vane after preliminary coating, as provided in an embodiment of the present invention. The determination of multiple wear areas of the test turbine guide vane after preliminary coating includes steps S41 to S43:

[0041] Step S41: Determine multiple wear point information of the test turbine guide vanes after preliminary coating based on the operating status data of the test turbine guide vanes during operation.

[0042] In some embodiments, a wear point determination model can be used to determine multiple wear point information of the test turbine guide vanes after preliminary coating. The wear point determination model is a deep neural network model. The input to the wear point determination model is the operating status data of the test turbine guide vanes during operation after preliminary coating, and the output of the wear point determination model is the multiple wear point information of the test turbine guide vanes after preliminary coating.

[0043] The multiple wear point information of the test turbine guide vanes after preliminary spraying is a description of the specific locations where microscopic wear occurred on the guide vane surface coating, determined by the wear point determination model after the test turbine guide vanes were initially sprayed and put into operation.

[0044] Wear point information includes the location of the wear point, the size of the wear point, the depth of the wear point, and the surface roughness of the wear point.

[0045] The operational data of the test turbine guide vanes after initial coating includes subtle signal differences on various parts of the guide vane surface caused by water flow impact, particle friction, and cavitation erosion. These differences are directly related to wear; for example, wear points can lead to localized stress concentration, vibration spectrum shifts, and fluctuations in water flow signals. This data forms the signal basis for identifying the location of wear.

[0046] Deep neural networks possess powerful feature learning and pattern recognition capabilities, enabling them to detect abnormal patterns or feature changes related to localized wear from complex operational data streams. Deep neural networks can perform in-depth analysis of operational data from test turbine guide vanes after initial coating. Through feature extraction and pattern recognition using multi-layer neural networks, the model can filter wear-related feature information from complex operational data. Based on this feature information, the model can accurately locate specific points of wear on the guide vane surface and infer key attributes such as the location, area, depth, and surface roughness of each wear point, ultimately determining information on multiple wear points.

[0047] In some embodiments, determining multiple wear point information of the test turbine guide vanes after preliminary coating based on the operating status data of the test turbine guide vanes during operation includes steps S411 to S413:

[0048] Step S411: Based on the operating status data of the test turbine guide vanes after the initial spraying, determine information on multiple abnormal data points.

[0049] In some embodiments, a deep neural network can be used to determine information about multiple anomalous data points.

[0050] Multiple abnormal data points are identified by a deep neural network in the running status data stream as discrete abnormal data records that deviate from the normal operating benchmark.

[0051] The abnormal data point information includes the timestamp of the abnormality, the duration of the abnormality, the specific value of the abnormal data, and the sensor channel identifier corresponding to the abnormal data.

[0052] Deep neural networks capture anomalies by performing time-series analysis on the operational status data of test turbine guide vanes after initial coating. The model's input layer receives continuous time-series data, while the hidden layers construct a probability distribution model of the guide vane's operational data under normal conditions. When new data is input, the neural network calculates the deviation distance between that data point and the normal pattern. The convolutional layers in the network can capture the trend of data changes over time. If the friction coefficient suddenly exceeds a preset confidence interval at a certain moment, or the vibration amplitude exhibits a nonlinear divergent trend, neurons will generate a high-intensity activation response. The model amplifies these minute signal fluctuations through nonlinear transformations and separates them from background noise. Utilizing its powerful feature extraction capabilities, the deep neural network can identify data segments that do not conform to the characteristics of fluid dynamics stability, such as pressure pulse spikes caused by turbulence resulting from localized coating peeling. Finally, the output layer marks these data points that are determined to be statistically significant, identifying them as multiple anomalous data points.

[0053] Step S412: Based on the information of the multiple abnormal data points, determine the abnormal intensity level, abnormal type, similarity with adjacent abnormal data points, impact index on guide vane operation, and probability of abnormal recurrence for each abnormal data point.

[0054] In some embodiments, a deep neural network can be used to determine the anomaly intensity level, anomaly type, similarity to neighboring anomaly data points, impact index on guide vane operation, and probability of anomaly recurrence for each anomalous data point.

[0055] The anomaly intensity level for each outlier data point is a numerical indicator output by a deep neural network, quantifying the degree to which the outlier data deviates from the normal baseline. The anomaly intensity level is used to determine the severity of the anomaly.

[0056] The anomaly type of each anomalous data point is a classification of the physical causes that caused the anomaly, determined by a deep neural network.

[0057] Anomaly types include coating peeling anomalies, coating cracking anomalies, and coating wear anomalies.

[0058] The similarity between each outlier data point and its neighboring outlier data points is an index determined by a deep neural network, representing the degree of proximity of the current outlier point to its temporal or spatial neighbors in terms of feature values.

[0059] The impact index of each abnormal data point on guide vane operation is a predicted numerical indicator determined by a deep neural network, representing the negative impact of the abnormal point on the overall hydraulic efficiency and safety of the guide vane.

[0060] The probability of an anomaly recurring is a statistical value determined by a deep neural network, representing the likelihood that the anomaly will recur in future operations.

[0061] The probability of abnormal recurrence of outlier data points can be used to predict coating failure trends.

[0062] Deep neural networks, with their powerful feature learning and association analysis capabilities, can accurately process multiple anomalous data points and output key evaluation results. By analyzing basic information such as the occurrence time, content description, and duration of anomalous data points from multiple dimensions, deep neural networks can automatically identify the characteristic patterns of different anomalies, thereby classifying anomaly types and quantifying their intensity levels. Simultaneously, deep neural networks can utilize data association mining to calculate the feature matching degree between a single anomaly and its adjacent anomalies, deriving a similarity index. Furthermore, the model can combine preset guide vane operation safety thresholds with multi-parameter coupling analysis to evaluate the impact index of anomalies on operation, and based on the temporal patterns and frequency distributions of anomaly occurrences, learn the probability patterns of their recurrence, ultimately completing anomaly assessment efficiently.

[0063] Step S413: Determine multiple wear point information based on the abnormal intensity level, abnormal type, similarity with adjacent abnormal data points, impact index on guide vane operation, and probability of abnormal recurrence for each abnormal data point.

[0064] In some embodiments, a deep neural network can be used to determine information about multiple wear points.

[0065] The deep neural network can receive parameters such as the anomaly intensity level, anomaly type, similarity to adjacent anomaly data points, impact index on guide vane operation, and probability of anomaly recurrence for each anomalous data point. These multi-dimensional data are then synchronously input into the feature fusion module and integrated through weighted calculation to form a unified feature vector. The deep neural network can match the anomaly type with preset wear formation rules and preliminarily estimate the degree of wear by combining the anomaly intensity level and impact index. The model can aggregate related data based on the similarity of adjacent anomalies to lock the wear area range, while referring to the probability of anomaly recurrence to correct the area boundary and depth value. Finally, through quantization mapping, the integrated feature data is transformed into specific information such as the location, area size, depth value, and surface roughness of the wear points, outputting multiple wear point information.

[0066] Step S42: Based on the information of multiple wear points on the test turbine guide vanes after the initial spraying, clustering is performed to obtain multiple clusters.

[0067] The clustering method used is K-means clustering. K-means clustering is an iterative clustering analysis algorithm used to divide a dataset into K distinct clusters, each containing data points with similar characteristics. The value of K can be predetermined manually.

[0068] Multiple clusters are the grouping results obtained by clustering multiple wear point information of the test turbine guide vanes after preliminary spraying using the K-means clustering algorithm. Each cluster contains multiple wear point information with similar characteristics. Wear points within the same cluster have high similarity in characteristics such as wear degree, location distribution, and operating conditions.

[0069] The process of clustering multiple wear points on the test turbine guide vanes after initial coating using the K-means clustering algorithm is as follows: First, K wear points are randomly selected from the multiple wear point data dataset as initial cluster centers. Next, for each wear point in the dataset, Euclidean distance is used to measure and calculate its comprehensive distance to these K initial cluster centers in terms of features such as location, area, depth, and surface roughness. The wear point is then assigned to the corresponding cluster based on the principle of closest proximity. After all wear points have been partitioned, the average values ​​of each feature of the wear point information within each cluster are recalculated to update the cluster centers of each cluster. This process of partitioning and updating cluster centers is repeated until the change in the position of the cluster centers is less than a preset threshold. At this point, the clustering process is considered to have converged, thus completing the K-means clustering.

[0070] Clustering can effectively integrate complex information on multiple wear points. The test turbine guide vanes after initial coating have a large number of wear points, and these points vary in location, area, depth, and surface roughness. Analyzing each individual wear point is not only inefficient but also makes it difficult to discover the overall distribution pattern of the wear. However, using the K-means clustering algorithm to group wear points with similar characteristics into multiple clusters greatly simplifies the data structure and reduces the complexity of subsequent analysis. Dividing the wear point information into multiple clusters allows for a visual representation of the different types and concentrated distribution of wear on the test turbine guide vane surface. Analysis of these clusters can quickly identify the core characteristics and high-incidence areas of different wear types.

[0071] Step S43: Based on the multiple clusters, determine the information of multiple wear areas of the test turbine guide vanes after preliminary spraying.

[0072] In some embodiments, a wear region segmentation model can be used to determine multiple wear region information of the test turbine guide vanes after preliminary coating. The wear region segmentation model is a deep neural network model. The input to the wear region segmentation model is the plurality of clusters, and the output of the wear region segmentation model is the multiple wear region information of the test turbine guide vanes after preliminary coating.

[0073] Wear area information is obtained by integrating and analyzing multiple clusters of wear points after clustering through a wear area segmentation model, and then defining the area on the guide vane surface.

[0074] The wear area information includes the boundary contour coordinates of the wear area, the average wear depth within the area, the dominant wear type, the wear development rate, the coating failure risk level, and the type of stress load on the area.

[0075] The types of loads include shear loads, impact loads, and alternating loads.

[0076] Multiple clusters are groups of wear points with similar characteristics, each cluster representing a common type of wear condition. These clusters can reflect the aggregation patterns of wear points in terms of spatial distribution and characteristics.

[0077] Deep neural networks can comprehensively analyze the features of multiple clusters and learn the spatial distribution patterns and common features of wear points within a cluster. By defining the boundary range of a cluster, a deep neural network can delineate the area containing wear points with related features within the same cluster and adjacent clusters as a wear region. The model first performs feature statistics on the wear point information within each cluster to obtain a set of wear features corresponding to each cluster, specifically including cluster-level features based on point features such as the spatial distribution trend of wear points within the cluster, average wear depth, and roughness distribution pattern of wear points. Subsequently, the model selects clusters that meet matching conditions, specifically those with continuous spatial distribution of adjacent clusters, cluster-level wear feature differences within a preset threshold range, and clusters located in the same structural partition or continuous operating area of ​​the guide vane. Based on the matching results, the model can integrate the qualified clusters into a continuous wear region, accurately define the boundary contour coordinates of this region, extract key attribute information of the region, and finally determine the multiple wear region information of the test turbine guide vane after preliminary spraying.

[0078] Step S5: Based on the information of multiple wear areas of the test turbine guide vanes after the initial spraying, generate secondary test spraying information for each wear area. The secondary test spraying information for each wear area includes graphene content and spraying thickness.

[0079] In some embodiments, Figure 4This is a schematic flowchart illustrating the process of generating secondary test spraying information for each wear area according to an embodiment of the present invention. The generation of secondary test spraying information for each wear area includes steps S51-S52:

[0080] Step S51: Based on the information of multiple wear areas of the test turbine guide vanes after the initial spraying, determine multiple representative wear areas, secondary test spraying information of each representative wear area, and spraying scheme difference.

[0081] In some embodiments, a wear region analysis model can be used to determine multiple representative wear regions, secondary test coating information for each representative wear region, and coating scheme differences. The wear region analysis model is a Transformer model. The input to the wear region analysis model is the wear region information of multiple wear regions of the test turbine guide vanes after the initial coating, and the output of the wear region analysis model is multiple representative wear regions, secondary test coating information for each representative wear region, and coating scheme differences.

[0082] The Transformer model is a deep learning model architecture primarily based on the self-attention mechanism. The core components of the Transformer model are the self-attention layer and the feedforward neural network layer. The self-attention mechanism allows the model to dynamically calculate the association weights between each element in the sequence and all other elements when processing sequential data, thereby capturing long-distance dependencies.

[0083] Multiple representative wear areas were selected from multiple wear areas of the test turbine guide vanes after initial coating using a wear area analysis model, and these areas exhibited typical characteristics.

[0084] Representative wear areas can reflect the core characteristics of different types of wear areas. By identifying multiple representative wear areas, it is possible to classify and integrate similar wear characteristics, avoiding redundant analysis of scattered wear points.

[0085] The secondary test coating information for each representative wear area is a set of process parameters determined by the wear area analysis model for each representative wear area, used for secondary test coating. This information includes the graphene content and coating thickness required for secondary coating in that area.

[0086] The coating scheme variability is an indicator of the degree of difference between secondary test coating information from different representative wear areas. It measures the compatibility of different representative areas with varying graphene content and coating thickness requirements.

[0087] The information on multiple wear areas of the test turbine guide vanes after initial spraying includes detailed characteristics of each wear area, which can reflect the differences in wear degree and extent of different wear areas.

[0088] The Transformer model, leveraging its powerful self-attention mechanism, can simultaneously focus on information from all wear regions and calculate their interrelationships. It can identify which regions are most representative, thus defining them as representative wear areas, and generate targeted secondary coating information based on the specific wear characteristics of each region. The model can adjust the graphene content and coating thickness according to the severity of wear. Furthermore, by comparing the coating parameters corresponding to different representative regions, the model can calculate the degree of difference in coating schemes.

[0089] Step S52: Based on the secondary test spraying information of each representative wear area and the difference of the spraying scheme, generate secondary test spraying information for other wear areas.

[0090] In some embodiments, a spraying information matching model can be used to generate secondary test spraying information for other wear areas. The spraying information matching model is a generative adversarial network (GAN). The inputs to the spraying information matching model are the secondary test spraying information for each representative wear area and the spraying scheme difference degree; the output of the spraying information matching model is the secondary test spraying information for other wear areas.

[0091] Generative Adversarial Networks (GANs) consist of two parts: a generator and a discriminator. The generator captures the data distribution and generates new data samples, while the discriminator distinguishes whether the input data is real or generated by the generator. Through adversarial training between the generator and discriminator, the GAN enables the generator to produce high-quality data that conforms to the target distribution.

[0092] Other wear areas are the wear areas on the test turbine guide vanes after initial coating, excluding the representative wear areas.

[0093] The secondary test spraying information for other wear areas is the spraying parameters for the remaining wear areas other than the representative area, output by the spraying information matching model. This secondary test spraying information for other wear areas includes the graphene content and spraying thickness of those areas.

[0094] The secondary test spraying information for each representative wear area provides the model with high-quality parameter samples. The difference in spraying schemes defines the reasonable range and diversity requirements of parameter changes, thereby providing the model with feature benchmarks and variation constraints for generating new parameters, so as to ensure that the parameters generated by the model are both in line with physical laws and have specificity.

[0095] In a generative adversarial network (GAN), the generator learns the patterns inherent in the secondary test spraying information and spraying scheme differences for each representative wear area. It then combines this with the correlation between the wear characteristics of other wear areas and the representative areas to generate secondary test spraying information adapted to other wear areas. The discriminator verifies the generated spraying information, determining whether it meets the corresponding wear area characteristic requirements. The generator can continuously optimize the generation results based on feedback from the discriminator, ultimately outputting reasonable secondary test spraying information for other wear areas.

[0096] Step S6: Based on the secondary test coating information of each wear area, perform secondary coating on the test turbine guide vanes after the initial coating, and obtain the operating status data of the test turbine guide vanes during operation after secondary coating.

[0097] The operating status data of the test turbine guide vanes after secondary coating are physical parameter records collected in the actual operating environment after the coating is applied according to the secondary test coating information for each wear area. These records are used to evaluate the performance of the secondary coating.

[0098] The operational status data of the test turbine guide vanes after secondary coating includes hydraulic efficiency maintenance index data during operation, stress distribution on the guide vane surface, vibration spectrum, and temperature field data of the coating area.

[0099] Step S7: Determine the target coating scheme based on the operating status data of the test turbine guide vanes after the secondary coating.

[0100] In some embodiments, Figure 5 This is a schematic flowchart illustrating the process of determining a target spraying scheme according to an embodiment of the present invention. The process of determining the target spraying scheme includes steps S71 to S74:

[0101] Step S71: Determine the wear information of each wear area after secondary test spraying based on the operating status data of the test turbine guide vanes after the secondary spraying.

[0102] In some embodiments, a wear information determination model can be used to determine the wear information of each wear area after secondary test coating. The wear information determination model is a deep neural network model. The input to the wear information determination model is the operating status data of the test turbine guide vanes after secondary coating, and the output of the wear information determination model is the wear information of each wear area after secondary test coating.

[0103] The wear information of each wear area after secondary spraying is an evaluation data of the wear resistance performance of each area after secondary spraying repair, determined by the wear information determination model.

[0104] The wear information after secondary spraying for each wear area includes the remaining coating thickness in the area, the change in surface roughness in the area, and the density of microcracks in the area.

[0105] The wear information after secondary testing of each wear area is used to quantify the coating effect.

[0106] The operational data of the test turbine guide vanes after the second coating process recorded the complete working process of the guide vanes and the actual wear of the coating. This data directly reflects the protective effect of the second coating scheme on each wear area.

[0107] Deep neural networks process input data by establishing an inverse mapping relationship between operating parameters and physical states. The input layer of a deep neural network receives operating state data after secondary coating, while the hidden layers analyze time-varying features in the data; for example, a slowing wear rate corresponds to improved wear resistance. The model can decompose and map global operating state data to specific spatial regions and infer the local state of each wear area by utilizing the differences in sensor responses at different locations. If the sensor signal for a certain region shows stable vibration and low frictional heat, the model can determine that the coating in that region is well maintained and calculate a higher remaining coating thickness. Conversely, if the signal exhibits high-frequency disturbances, the model infers an increase in surface roughness in that region. Through nonlinear regression analysis, deep neural networks can transform indirect signal features into direct physical quantitative indicators. Finally, the output layer outputs wear information for each wear area after secondary coating testing.

[0108] Step S72: Construct a secondary spraying pattern. The secondary spraying pattern includes multiple wear areas and multiple edges between the multiple wear areas. The node features of each wear area are the wear information after secondary test spraying of each wear area and the secondary test spraying information of each wear area.

[0109] A graph is a data structure used to represent entities and their relationships. A graph consists of vertices and edges. A secondary coating graph is a graph that presents information related to wear areas in a graphical form. Nodes in a secondary coating graph correspond to multiple wear areas, and node features include wear information after a secondary test coating and secondary test coating information for each wear area.

[0110] Step S73: Process the secondary spraying pattern based on graph neural network to obtain secondary target spraying information for each wear area.

[0111] Graph Neural Networks (GNNs) are deep learning models capable of processing graphs. Through message passing mechanisms, GNNs enable nodes in a graph to aggregate information from their neighbors to update their feature representations. GNNs can simultaneously capture local node features and global graph topology information, and are suitable for tasks such as node classification, link prediction, and graph classification. The input to the GNN is the secondary spraying map, and the output is the secondary target spraying information for each wear area.

[0112] The secondary target coating information for each wear area is determined by a graph neural network based on the actual wear degree, wear type, and other local features of a single wear area on the guide vane, resulting in precise protective coating parameters. The secondary target coating information for each wear area includes the target graphene content and target coating thickness that achieve the best wear-resistant protection effect in that area.

[0113] The core of secondary target spraying information is to focus on maximizing the wear resistance requirements of the wear area, thereby ensuring that the wear resistance protection effect in that area reaches the optimal level.

[0114] In the secondary coating map, nodes correspond to different wear regions, and edges represent the adjacency or functional relationships between different wear regions. Each wear region node possesses rich node features, such as wear information after the secondary test coating of that region, as well as information on secondary test coatings previously applied to that region. Graph neural networks can process this complex graph data, updating and enriching the feature representation of each node through information transfer and aggregation mechanisms between neighboring nodes. This representation can capture the mutual influence and synergistic effects between regions. The secondary coating map provides structured information about wear regions and their relationships, while graph neural networks have the ability to learn complex dependencies between nodes from the graph and can generate better secondary target coating information for each wear region based on the learned global and local information.

[0115] Graph neural networks (Graph Neural Networks) process the secondary coating pattern and utilize a message passing mechanism to allow each node in a wear area to receive feature information from its neighboring nodes, achieving information sharing and fusion among nodes. Graph Neural Networks can not only learn the characteristics of each wear area itself but also capture the correlations between different wear areas. Based on this comprehensive information, the Graph Neural Network optimizes and adjusts the coating scheme for each wear area, ultimately outputting the secondary target coating information for each wear area.

[0116] Step S74: Determine the target coating scheme based on the preliminary coating thickness information and the secondary target coating information of each wear area.

[0117] In some embodiments, a spraying scheme determination model can be used to determine the target spraying scheme. The spraying scheme determination model is a deep neural network model. The inputs to the spraying scheme determination model are the initial spraying thickness information and the secondary target spraying information for each wear area; the output of the spraying scheme determination model is the target spraying scheme.

[0118] The target coating scheme is a final execution plan used to guide the coating of graphene ceramic gold coating on the turbine guide vanes of the product line. It is formed by integrating the initial coating thickness information and the secondary target coating information of all wear areas and then optimizing it globally through a deep neural network. The target coating scheme includes graphene ceramic gold coating path planning, spray gun movement speed at each path point, and coating flow rate setting at each path point.

[0119] The initial coating thickness information provides the basic protection parameters for the entire guide vane, while the secondary target coating information for each wear area clarifies the enhanced protection requirements for vulnerable areas. Both cover the complete needs of the entire guide vane area from basic protection to key reinforcement, providing all-element input for the model to generate a target coating scheme that balances efficiency and protection effect, ensuring the completeness and relevance of the input data.

[0120] After the input layer of the deep neural network receives the initial coating thickness information and the secondary target coating information for each wear area, the hidden layer performs fusion operations on the input data. The hidden layer matches and integrates the secondary target coating information for each wear area with the initial coating thickness information to construct the final 3D morphology model of the coating. The model fully considers the kinematic constraints of the painting robot, transforming the ideal coating thickness distribution into an executable robotic arm trajectory. The deep neural network balances painting efficiency and coating quality through optimization algorithms. For example, it uses a fast movement mode in areas corresponding to the initial coating thickness information, while automatically reducing the movement speed and increasing the spray flow rate in areas corresponding to the secondary target coating information of each wear area. The model also solves the transition problem between the base layer and the reinforcement layer through smoothing processing to prevent the formation of a step effect. Finally, the output layer decodes and generates a target coating scheme containing specific path, speed, and flow rate parameters.

[0121] Step S8: Based on the target spraying scheme, control the spraying of graphene ceramic coating on the turbine guide vanes of the product line.

[0122] Once the target coating scheme is determined, the turbine guide vanes of the product line are coated with a graphene ceramic gold coating that meets the requirements by executing the target coating scheme, thereby achieving effective protection of the guide vanes and extending their service life.

[0123] Based on the same inventive concept Figure 6A schematic diagram of a turbine guide vane spraying control system based on a neural network model is provided in this embodiment of the invention. The turbine guide vane spraying control system based on a neural network model includes:

[0124] Module 91 is used to acquire historical operating status data of the turbine guide vanes during operation;

[0125] The preliminary coating thickness determination module 92 is used to determine the preliminary coating thickness information based on the historical operating status data of the turbine guide vanes during operation, using an operation analysis model, wherein the operation analysis model is a deep neural network model.

[0126] The preliminary coating control module 93 is used to control the coating robot to perform preliminary coating of graphene ceramic gold coating on the test turbine guide vane based on the preliminary coating thickness information, and to obtain the operating status data of the test turbine guide vane during operation after preliminary coating.

[0127] Wear area determination module 94 is used to determine multiple wear area information of the test turbine guide vane after preliminary coating based on the operating status data of the test turbine guide vane after preliminary coating.

[0128] The secondary test coating information generation module 95 is used to generate secondary test coating information for each wear area based on the multiple wear area information of the test turbine guide vane after the initial coating. The secondary test coating information for each wear area includes graphene content and coating thickness.

[0129] The secondary spraying control module 96 is used to perform secondary spraying on the test turbine guide vanes after the initial spraying based on the secondary test spraying information of each wear area, and to obtain the operating status data of the test turbine guide vanes during operation after secondary spraying.

[0130] The target spraying scheme determination module 97 is used to determine the target spraying scheme based on the operating status data of the test turbine guide vanes after the secondary spraying.

[0131] The spraying control module 98 is used to control the spraying of graphene ceramic coating on the turbine guide vanes of the product line based on the target spraying scheme.

[0132] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0133] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A guide vane spraying control method based on a neural network model, characterized by, The method comprises: acquiring historical operation state data of the guide vane of the water turbine when the guide vane is in operation; determining preliminary spraying thickness information by using an operation analysis model based on the historical operation state data of the guide vane of the water turbine when the guide vane is in operation, the operation analysis model being a deep neural network model; controlling a spraying robot to perform preliminary spraying of a graphene ceramic gold coating on the test guide vane of the water turbine based on the preliminary spraying thickness information, and acquiring operation state data of the test guide vane of the water turbine when the test guide vane is in operation after the preliminary spraying; determining multiple wear area information of the test guide vane of the water turbine after the preliminary spraying based on the operation state data of the test guide vane of the water turbine when the test guide vane is in operation after the preliminary spraying, the determination comprising: determining multiple wear point information of the test guide vane of the water turbine after the preliminary spraying based on the operation state data of the test guide vane of the water turbine when the test guide vane is in operation after the preliminary spraying; performing clustering based on the multiple wear point information of the test guide vane of the water turbine after the preliminary spraying to obtain multiple clusters; determining the multiple wear area information of the test guide vane of the water turbine after the preliminary spraying based on the multiple clusters; generating secondary test spraying information of each wear area based on the multiple wear area information of the test guide vane of the water turbine after the preliminary spraying, the secondary test spraying information of each wear area comprising graphene content and spraying thickness; performing secondary spraying on the test guide vane of the water turbine after the preliminary spraying based on the secondary test spraying information of each wear area, and acquiring operation state data of the test guide vane of the water turbine when the test guide vane is in operation after the secondary spraying; determining a target spraying scheme based on the operation state data of the test guide vane of the water turbine when the test guide vane is in operation after the secondary spraying, the determination comprising: determining wear information of each wear area after the secondary test spraying based on the operation state data of the test guide vane of the water turbine when the test guide vane is in operation after the secondary spraying; constructing a secondary spraying graph, the secondary spraying graph comprising multiple wear areas and multiple edges between the multiple wear areas, and the node feature of each wear area being the wear information of each wear area after the secondary test spraying and the secondary test spraying information of each wear area; processing the secondary spraying graph based on a graph neural network to obtain secondary target spraying information of each wear area; determining the target spraying scheme based on the preliminary spraying thickness information and the secondary target spraying information of each wear area; and controlling spraying of the graphene ceramic gold coating on the guide vane of the product line based on the target spraying scheme. 2.The water turbine guide vane spraying control method based on a neural network model according to claim 1, wherein, The generation of the secondary test spraying information of each wear area based on the multiple wear area information of the test guide vane of the water turbine after the preliminary spraying comprises: determining multiple representative wear areas, secondary test spraying information of each representative wear area, and spraying scheme difference degree based on the multiple wear area information of the test guide vane of the water turbine after the preliminary spraying. Generate secondary test spraying information of other wear areas based on the secondary test spraying information of each representative wear area and the spraying scheme difference degree.

3. A guide vane spraying control system for a hydraulic turbine based on a neural network model, characterized by, Comprise: An acquisition module is configured to acquire historical operating state data of the guide vanes of the water turbine; A preliminary spraying thickness determination module is configured to determine preliminary spraying thickness information based on the historical operating state data of the guide vanes of the water turbine using an operating analysis model, which is a deep neural network model; A preliminary spraying control module is configured to control a spraying robot to perform preliminary spraying of graphene molybdenum coating on the test guide vanes of the water turbine based on the preliminary spraying thickness information, and to acquire operating state data of the test guide vanes of the water turbine after preliminary spraying; A wear area determination module is configured to determine multiple wear area information of the test guide vanes of the water turbine after preliminary spraying based on the operating state data of the test guide vanes of the water turbine after preliminary spraying, and to further determine: Determine multiple wear point information of the test guide vanes of the water turbine after preliminary spraying based on the operating state data of the test guide vanes of the water turbine after preliminary spraying; Cluster the multiple wear point information of the test guide vanes of the water turbine after preliminary spraying to obtain multiple clusters; Determine multiple wear area information of the test guide vanes of the water turbine after preliminary spraying based on the multiple clusters; A secondary test spraying information generation module is configured to generate secondary test spraying information of each wear area based on the multiple wear area information of the test guide vanes of the water turbine after preliminary spraying, wherein the secondary test spraying information of each wear area comprises graphene content and spraying thickness; A secondary spraying control module is configured to perform secondary spraying on the test guide vanes of the water turbine after preliminary spraying based on the secondary test spraying information of each wear area, and to acquire operating state data of the test guide vanes of the water turbine after secondary spraying; A target spraying scheme determination module is configured to determine a target spraying scheme based on the operating state data of the test guide vanes of the water turbine after secondary spraying, and to further determine: Determine wear information of each wear area after secondary test spraying based on the operating state data of the test guide vanes of the water turbine after secondary spraying; Construct a secondary spraying graph, wherein the secondary spraying graph comprises multiple wear areas and multiple edges between the multiple wear areas, and the node features of each wear area are the wear information of each wear area after secondary test spraying and the secondary test spraying information of each wear area; Process the secondary spraying graph based on a graph neural network to obtain secondary target spraying information of each wear area; Determine a target spraying scheme based on the preliminary spraying thickness information and the secondary target spraying information of each wear area; A spraying control module is configured to perform spraying control of graphene molybdenum coating on the guide vanes of the water turbine of the product line based on the target spraying scheme.

4. The guide vane spraying control system for a hydraulic turbine based on a neural network model according to claim 3, wherein The secondary test spraying information generation module is further configured to: determine a plurality of representative wear areas based on the plurality of wear area information of the test guide vane after the preliminary spraying, secondary spraying information of each representative wear area, and a spraying scheme difference degree; generate secondary spraying information of other wear areas based on the secondary spraying information of each representative wear area and the spraying scheme difference degree.

5. An electronic device, comprising: comprise: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the water turbine guide vane spraying control method based on the neural network model as claimed in any one of claims 1 to 2.

6. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the water turbine guide vane spraying control method based on the neural network model as claimed in any one of claims 1 to 2.

Images