Data analysis based method and system for synergistic optimization of graphene ceramic coating formulation for water turbines

Abstract

The application provides a water turbine graphene ceramic coating formula synergistic optimization method and system based on data analysis, and relates to the water turbine coating optimization technical field.The method comprises the following steps: obtaining historical flow field data and an image of a water turbine flow part;generating abrasion distribution information of the water turbine flow part based on the historical flow field data and the image of the water turbine flow part; determining target coating formula information of a test formula adjustment region based on abrasion test data of the same test sample of the water turbine flow part under each preliminary coating formula information; and determining target coating formula information of each remaining formula adjustment region based on the target coating formula information of the test formula adjustment region.The method can accurately determine the target formula of the graphene ceramic coating suitable for each region of the water turbine flow part, so as to realize global formula synergistic optimization.

Description

Technical Field

[0001] This invention relates to the field of turbine coating optimization technology, specifically to a method and system for collaborative optimization of turbine graphene ceramic coating formulations based on data analysis. Background Technology

[0002] Hydropower, as a crucial component of green energy, relies heavily on the operational efficiency and reliability of its core equipment, the turbine. Turbine components are subjected to harsh conditions of high-speed water flow erosion, cavitation, and sediment abrasion, leading to surface erosion damage and consequently, decreased equipment performance and increased maintenance costs. To improve the wear resistance of these components, applying graphene ceramic coatings has become an important industry practice. However, traditional methods for determining turbine coating formulations rely heavily on experience and standardized experimental ratios, lacking personalized optimization for specific operating conditions. This traditional approach has significant limitations. First, the internal flow field of a turbine is complex, with significant differences in erosion intensity and wear mechanisms across different parts, making it difficult for a single or empirically based coating formulation to achieve ideal protection across the entire flow surface. Second, traditional offline experiments often fail to fully simulate the actual erosion state of components under complex flow fields, resulting in discrepancies between experimental data and practical applications, and insufficient basis for site selection and formulation determination. Furthermore, existing formulation adjustments often employ a standardized approach, lacking in-depth analysis of the erosion correlation between different areas of the flow-through components, making it difficult to achieve synergistic optimization of the formulation across different damaged areas. This not only limits further improvements in coating protective performance but also leads to waste of material resources and low maintenance efficiency, failing to meet the demands of modern hydroelectric power plants for precise and efficient equipment management.

[0003] Therefore, how to accurately determine the target formulation of graphene ceramic coating for each region of the turbine's flow-through components in order to achieve global formulation synergistic optimization is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to accurately determine the target formulation of graphene ceramic coating for each region of the turbine flow-through components in order to achieve global formulation synergistic optimization.

[0005] According to a first aspect, the present invention provides a method for collaborative optimization of graphene ceramic coating formulations for water turbines based on data analysis, comprising: acquiring historical flow field data and images of water turbine flow components; generating abrasion distribution information of water turbine flow components based on the historical flow field data and images of water turbine flow components; determining, based on the abrasion distribution information and images of water turbine flow components, a test formulation adjustment area and multiple other formulation adjustment areas of the water turbine flow components; and determining, based on the abrasion distribution information and images of the water turbine flow components, a test formulation adjustment area and multiple other formulation adjustment areas of the water turbine flow components; and determining, based on the abrasion distribution information of the test formulation adjustment area, a test formulation adjustment area and multiple other formulation adjustment areas of the water turbine flow components. Multiple identical test samples and multiple preliminary coating formulations for corresponding turbine flow components are identified, with each preliminary coating formulation having a different graphene content and ceramic powder ratio. Abrasion test data of the identical test samples of the turbine flow components under each preliminary coating formulation is obtained. Based on the abrasion test data of the identical test samples of the turbine flow components under each preliminary coating formulation, the target coating formulation for the tested area to be adjusted is determined. Based on the target coating formulation for the tested area to be adjusted, the target coating formulation for each of the remaining areas to be adjusted is determined.

[0006] In one possible implementation, determining the test formula adjustment area and multiple other formula adjustment areas of the turbine flow components based on the erosion distribution information and images of the turbine flow components includes: determining multiple erosion contour lines of the turbine flow components based on the erosion distribution information and images of the turbine flow components; determining boundary point information of multiple candidate formula adjustment areas based on the erosion distribution information and multiple erosion contour lines of the turbine flow components; determining multiple candidate formula adjustment area information based on the erosion distribution information and boundary point information of the multiple candidate formula adjustment areas; clustering the multiple candidate formula adjustment area information to obtain multiple clusters; and determining the test formula adjustment area and multiple other formula adjustment areas of the turbine flow components based on the multiple clusters.

[0007] In one possible implementation, determining the target coating formula information for each remaining region to be adjusted based on the target coating formula information of the test region to be adjusted includes: generating multiple simulated coating formula information for each remaining region to be adjusted based on the target coating formula information of the test region to be adjusted; constructing a coating formula simulation map, which includes multiple nodes of the remaining regions to be adjusted and edges between the multiple remaining regions to be adjusted, wherein the node features of each node of the region to be adjusted include abrasion distribution information of the remaining regions to be adjusted, multiple simulated coating formula information for each remaining region to be adjusted, abrasion distribution information of the test region to be adjusted, and target coating formula information of the test region to be adjusted; and processing the coating formula simulation map based on a coating formula optimization model to obtain the target coating formula information.

[0008] In one possible implementation, the input to the coating formulation optimization model is the coating formulation simulation map, and the output of the coating formulation optimization model is the target coating formulation information.

[0009] According to a second aspect, the present invention provides a data analysis-based collaborative optimization system for graphene ceramic coating formulations of water turbines, comprising: an acquisition module for acquiring historical flow field data and images of water turbine flow components; an abrasion distribution information generation module for generating abrasion distribution information of the water turbine flow components based on the historical flow field data and images of the water turbine flow components; an adjustment area determination module for determining a test formulation adjustment area and multiple other formulation adjustment areas of the water turbine flow components based on the abrasion distribution information and images of the water turbine flow components; and a preliminary formulation and sample determination module for determining the abrasion distribution information of the test formulation adjustment area. The system identifies multiple identical test samples and multiple preliminary coating formulations for the corresponding turbine flow components, with each preliminary coating formulation having a different graphene content and ceramic powder ratio. A wear test data acquisition module acquires wear test data for the identical test samples of the turbine flow components under each preliminary coating formulation. A test area target formulation determination module determines the target coating formulation for the test area requiring formulation adjustment based on the wear test data of the identical test samples of the turbine flow components under each preliminary coating formulation. A remaining area target formulation determination module determines the target coating formulation for each of the remaining areas requiring formulation adjustment based on the target coating formulation for the test area requiring formulation adjustment.

[0010] In one possible implementation, the adjustment area determination module is further configured to: determine multiple erosion contour lines of the turbine flow-through component based on the erosion distribution information and the image of the turbine flow-through component; determine boundary point information of multiple candidate formula adjustment areas based on the erosion distribution information and the multiple erosion contour lines of the turbine flow-through component; determine multiple candidate formula adjustment area information based on the erosion distribution information and the boundary point information of the multiple candidate formula adjustment areas; perform clustering based on the multiple candidate formula adjustment area information to obtain multiple clusters; and determine the test formula adjustment area and multiple other formula adjustment areas of the turbine flow-through component based on the multiple clusters.

[0011] In one possible implementation, the remaining region target formulation determination module is further configured to: generate multiple simulated coating formulation information for each remaining region to be adjusted based on the target coating formulation information of the test region to be adjusted; construct a coating formulation simulation map, which includes multiple nodes of the remaining regions to be adjusted and edges between the multiple remaining regions to be adjusted, wherein the node features of each node of the region to be adjusted include abrasion distribution information of the remaining regions to be adjusted, multiple simulated coating formulation information of each remaining region to be adjusted, abrasion distribution information of the test region to be adjusted, and target coating formulation information of the test region to be adjusted; and process the coating formulation simulation map based on the coating formulation optimization model to obtain target coating formulation information.

[0012] In one possible implementation, the input to the coating formulation optimization model is the coating formulation simulation map, and the output of the coating formulation optimization model is the target coating formulation information.

[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 flow field data and images of turbine flow components; generating erosion distribution information of turbine flow components based on the historical flow field data and images of turbine flow components; and determining, based on the erosion distribution information and images of turbine flow components, a test formulation adjustment area and multiple other formulations to be adjusted for the turbine flow components. Adjustment area; based on the abrasion distribution information of the test area to be adjusted, determine multiple identical test samples and multiple preliminary coating formulations for the corresponding turbine flow components, with different graphene content and ceramic powder ratios in each preliminary coating formulation; obtain abrasion test data of the identical test samples of the turbine flow components under each preliminary coating formulation; based on the abrasion test data of the identical test samples of the turbine flow components under each preliminary coating formulation, determine the target coating formulation information for the test area to be adjusted; based on the target coating formulation information of the test area to be adjusted, determine the target coating formulation information for each of the remaining areas to be adjusted.

[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 data analysis-based collaborative optimization method for graphene ceramic coating formulations of water turbines. The method includes: acquiring historical flow field data and images of water turbine flow components; generating abrasion distribution information of water turbine flow components based on the historical flow field data and the images of water turbine flow components; and determining, based on the abrasion distribution information and the images of water turbine flow components, a test formulation adjustment area and multiple other formulation adjustment areas of the water turbine flow components. Based on the abrasion distribution information of the test area to be adjusted, multiple identical test samples and multiple preliminary coating formulations of the corresponding turbine flow components are determined, with different graphene content and ceramic powder ratios in each preliminary coating formulation. Abrasion test data of the identical test samples of the turbine flow components under each preliminary coating formulation are obtained. Based on the abrasion test data of the identical test samples of the turbine flow components under each preliminary coating formulation, the target coating formulation information of the test area to be adjusted is determined. Based on the target coating formulation information of the test area to be adjusted, the target coating formulation information of each remaining area to be adjusted is determined.

[0015] This invention provides a data analysis-based method and system for collaborative optimization of graphene ceramic coating formulations for water turbines. The method includes acquiring historical flow field data and images of the water turbine's flow components; generating abrasion distribution information of the water turbine's flow components based on the historical flow field data and the images; determining, based on the abrasion distribution information and the images, a test area for formulation adjustment and multiple other areas for formulation adjustment of the water turbine's flow components; and determining, based on the abrasion distribution information of the test area for formulation adjustment, multiple identical test samples and multiple other test samples of the corresponding water turbine flow components based on the abrasion distribution information of the test area for formulation adjustment. The method involves several steps: First, obtaining preliminary coating formulation information, where the graphene content and ceramic powder ratio differ in each formulation. Second, acquiring abrasion test data for identical test samples of the turbine's flow components under each preliminary coating formulation. Third, determining the target coating formulation for the tested area requiring formulation adjustment based on the target coating formulation for the tested area. Finally, determining the target coating formulation for each of the remaining areas requiring formulation adjustment based on the target coating formulation for the tested area. This method can accurately determine the appropriate graphene ceramic coating target formulation for each region of the turbine's flow components to achieve global formulation synergistic optimization. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a data analysis-based collaborative optimization method for a graphene ceramic coating formulation for a water turbine, provided as an embodiment of the present invention;

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

[0018] Figure 3 This is a flowchart illustrating a process for determining the test formula adjustment area and multiple other formula adjustment areas of a turbine flow component, as provided in an embodiment of the present invention.

[0019] Figure 4 This is a schematic diagram of a process for determining the target coating formulation information for each remaining region to be adjusted, provided by an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of a data analysis-based collaborative optimization system for graphene ceramic coating formulations in water turbines, provided as an embodiment of the present invention. Detailed Implementation

[0021] 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.

[0022] In this embodiment of the invention, the following are provided: Figure 1 The method shown is a collaborative optimization method for the formulation of graphene ceramic coating for water turbines based on data analysis. The method includes steps S1 to S7:

[0023] Step S1: Obtain historical flow field data and images of the turbine's flow passage components.

[0024] The flow passage components of a water turbine are the collection of components through which fluid flows during the turbine's operation. These components may include the guide vanes, runner, spiral casing, and draft tube. The flow passage components are the core carriers of energy conversion in a water turbine and directly withstand fluid impact, friction, and corrosion. Figure 2 This is a schematic diagram of a water turbine flow passage component provided in an embodiment of the present invention.

[0025] Historical flow field data of turbine flow passage components are fluid dynamic state data collected from turbine flow passage components during past operating cycles. The historical flow field data of turbine flow passage components includes dynamic sequences of water flow velocity vector, pressure distribution, turbulence intensity, shear stress distribution, etc., changing over time.

[0026] Images of the flow-through components of a water turbine are high-resolution visual images obtained by taking pictures of the surface of the flow-through components of a water turbine using an industrial camera.

[0027] Images of turbine flow passage components can show the wear morphology of the component surface, the geometric features of material spalling pits, and can intuitively reflect the physical state of damage to turbine flow passage components.

[0028] Step S2: Based on the historical flow field data of the turbine flow passage components and the image of the turbine flow passage components, generate erosion distribution information of the turbine flow passage components.

[0029] In some embodiments, an erosion distribution analysis model can be used to generate erosion distribution information of the turbine's flow passage components. The erosion distribution analysis model is a deep neural network. The inputs to the erosion distribution analysis model are historical flow field data of the turbine's flow passage components and images of the turbine's flow passage components; the output of the erosion distribution analysis model is the erosion distribution information of the turbine's flow passage components.

[0030] Deep neural network models include deep neural networks (DNNs). A deep neural network is a machine learning model composed of multiple layers of stacked neurons, including an input layer, hidden layers, and an output layer. Hidden layers can fit complex mapping relationships using non-linear activation functions. Deep neural networks can automatically extract high-level features from input data.

[0031] The erosion distribution information of the turbine's flow passage components describes the degree of erosion at different locations on the surface of these components. This information includes the coordinates of the erosion location, the erosion depth at each location, the erosion area, the erosion rate, and the surface roughness coefficient.

[0032] Historical flow field data of turbine flow passage components reveals the scouring force and stress concentration patterns of the fluid on the component surface. This historical flow field data directly affects the occurrence and development of erosion. Images of turbine flow passage components present the appearance characteristics of erosion traces already formed on the surface of the components and can intuitively reflect the actual state of erosion.

[0033] Deep neural networks, through nonlinear activation functions in their hidden layers, can perform in-depth analysis of historical flow field data of input turbine components, thereby identifying characteristic regions in the flow field prone to cavitation erosion or sediment wear. The deep neural network maps the historical flow field data of the turbine components to a high-dimensional feature space and analyzes the correspondence between fluid energy loss and surface stress loads on the components. Simultaneously, it extracts visual damage features such as crack orientation and hole depth from images of the turbine components. Based on learned physical laws of abrasion, the deep neural network can determine the material damage patterns under different flow velocity ranges and pressure intensities, thereby calculating the abrasion depth and abrasion tendency of each grid cell on the component surface, ultimately accurately determining the abrasion distribution information of the turbine components covering the entire component surface.

[0034] Step S3: Based on the erosion distribution information of the turbine flow-through components and the image of the turbine flow-through components, determine the test formula adjustment area of ​​the turbine flow-through components and several other formula adjustment areas.

[0035] In some embodiments, Figure 3This is a schematic flowchart illustrating the process of determining the test formula adjustment area and multiple other formula adjustment areas of a turbine flow-through component according to an embodiment of the present invention. The determination of the test formula adjustment area and multiple other formula adjustment areas of the turbine flow-through component includes steps S31 to S35:

[0036] Step S31: Based on the erosion distribution information of the turbine flow-through components and the image of the turbine flow-through components, determine multiple erosion contour lines of the turbine flow-through components.

[0037] In some embodiments, an erosion contour determination model can be used to determine multiple erosion contours of the turbine flow passage components. The erosion contour determination model is a convolutional neural network model. The input to the erosion contour determination model is the erosion distribution information of the turbine flow passage components and an image of the turbine flow passage components; the output of the erosion contour determination model is multiple erosion contours of the turbine flow passage components.

[0038] Convolutional Neural Network (CNN) models are a type of neural network architecture used to process data with a grid structure. CNNs effectively extract local spatial features of the data by using a sliding window operation on the input data through convolutional kernels. A CNN consists of convolutional layers, pooling layers, and fully connected layers.

[0039] Multiple erosion contour lines of the turbine's flow passage components are closed curves formed by connecting points with the same degree of erosion on the surface of the turbine's flow passage components.

[0040] Multiple erosion contour lines on the flow passage components of a water turbine can clearly delineate the boundaries of regions with different degrees of erosion, with each contour line corresponding to a fixed erosion depth value.

[0041] The abrasion distribution information of the turbine's flow-through components provides a detailed numerical field, while the images of these components offer visual texture of their physical edges. The numerical gradients in the abrasion distribution information determine the orientation of the contour lines, while the color difference and morphological features in the images help the model verify the accuracy of these numerical gradients, ensuring that the abrasion contour lines determined by the model accurately reflect the damaged boundaries of the component surface.

[0042] Convolutional neural networks (CNNs) leverage their powerful feature extraction capabilities to spatially filter the erosion distribution information of turbine flow components, thereby identifying regions with significant rate of change in the numerical field. The CNN converts the input erosion distribution information of the turbine flow components into a three-dimensional grayscale matrix, then uses convolutional kernels of different scales to capture abrupt changes in erosion depth. Subsequently, the model matches the extracted numerical gradient features with the physical concave edges identified from the turbine flow component images. By classifying pixels through activation functions, the CNN determines which coordinate points have erosion indices within the same preset threshold range and uses a contour tracking algorithm to connect these points with the same indices into a smooth geometric curve. The CNN repeats this process for different erosion depth gradients, thereby globally determining multiple erosion contour lines reflecting the hierarchical distribution of erosion severity for the turbine flow components.

[0043] Step S32: Based on the erosion distribution information of the turbine flow-through components and the multiple erosion contour lines of the turbine flow-through components, determine the boundary point information of multiple candidate formula adjustment areas.

[0044] In some embodiments, a boundary point determination model can be used to determine the boundary point information of multiple candidate formulation adjustment regions. The boundary point determination model is a deep neural network model. The input to the boundary point determination model is the erosion distribution information of the turbine flow components and multiple erosion contour lines of the turbine flow components; the output of the boundary point determination model is the boundary point information of multiple candidate formulation adjustment regions.

[0045] The boundary point information of multiple candidate formulation adjustment regions is a set of coordinate data of key vertices at the edges of various regions that may require coating formulation optimization. The boundary point information of each candidate formulation adjustment region includes the position coordinates of each vertex of the candidate region, the local curvature change value, the erosion contour line number to which the boundary point belongs, and the erosion depth value at the boundary point.

[0046] Boundary point information of multiple candidate formulation adjustment areas can be used to accurately define the spatial range and geometric contour of each formulation adjustment area, providing a spatial positioning basis for subsequent area division and zonal adaptation of coating formulations. At the same time, by using the local curvature change value, abrasion depth and contour line number information carried by the boundary points, the transition characteristics of each area can be accurately identified, thereby avoiding abrupt changes or discontinuities in the coating formulation of adjacent areas, and ensuring the continuity and rationality of the overall coating protection of the turbine flow components.

[0047] The erosion distribution information of the turbine's flow-through components can clearly define the damage level in each candidate formulation adjustment area and identify the areas requiring formulation adjustments. Multiple erosion contour lines of the turbine's flow-through components can present the macroscopic distribution trend of damage intensity, thus providing a geometric reference for areas requiring formulation adjustments. The erosion distribution information and multiple erosion contour lines enable the model to accurately locate the inflection points of drastic changes in erosion intensity.

[0048] Based on the erosion distribution information and multiple erosion contour lines of turbine flow components, deep neural networks can model the spatial correlation of erosion intensity in hidden layers. By performing statistical analysis on the erosion distribution information of turbine flow components, deep neural networks can identify connected regions where the erosion depth exceeds a safe threshold. The model can determine the geometric complexity of connected regions by combining the topological structure of multiple erosion contour lines of turbine flow components. Then, the model can extract key features of inflection points, intersection points, and extreme points in the corresponding regions. Based on the learned region partitioning logic, the model can locate transition zones with gentle gradient changes around the high-erosion center region and determine the coordinate nodes within this range that can be closed to form a complete region. Through a logistic regression layer, the deep neural network can perform probabilistic screening on each candidate node to retain vertices that can effectively enclose the high-erosion region and satisfy the geometric constraints of the components, ultimately obtaining the boundary point information of multiple candidate formulation adjustment regions.

[0049] Step S33: Determine multiple candidate formula adjustment area information based on the erosion distribution information of the turbine flow components and the boundary point information of the multiple candidate formula adjustment areas.

[0050] In some embodiments, an adjustment region determination model can be used to determine information on multiple candidate formulation adjustment regions. The adjustment region determination model is a deep neural network model. The inputs to the adjustment region determination model are the erosion distribution information of the turbine flow components and the boundary point information of the multiple candidate formulation adjustment regions; the output of the adjustment region determination model is the information on the multiple candidate formulation adjustment regions.

[0051] The multiple candidate formulation adjustment areas information refers to the descriptive information of several specific locations on the surface of the turbine's flow-through components that require formulation optimization. Each candidate formulation adjustment area includes the boundary contour coordinates of the area, the importance level of the component's functional part to which the area belongs, the average erosion depth within the area, the maximum erosion depth within the area, the area's erosion risk assessment level, and the area's formulation optimization priority index.

[0052] Boundary point information for multiple candidate formulation adjustment regions can clearly define the closed spatial range and geometric boundaries of each candidate region, as well as the spatial distribution and contour morphology of each candidate formulation adjustment region. The erosion distribution information of the turbine's flow-through components includes the numerical distribution of erosion across the entire surface of the components, enabling the model to quantitatively describe the erosion characteristics within the bounded area of ​​the boundary points. By using the spatial range of the boundary point information as a statistical unit, and combining it with the erosion distribution information to statistically analyze the erosion characteristics within this range, the model can determine the candidate formulation adjustment regions.

[0053] Deep neural networks can perform geometric reconstruction based on the boundary point information of multiple candidate formulation adjustment regions. The model uses interpolation algorithms and surface fitting techniques to connect isolated vertex coordinates into closed polygonal regions. The deep neural network then uses a perceptron layer to analyze the coverage of these polygons on the surface of the turbine's flow-through components and performs spatial mapping matching with the input abrasion distribution information of the turbine's flow-through components. The model can calculate all abrasion values ​​contained within each closed polygon and extract the damage statistical features within that range. The model can perform a rationality check on the generated regions to eliminate invalid regions that are too small or have excessively different abrasion degrees, while merging adjacent regions with similar abrasion degrees, ultimately determining multiple candidate formulation adjustment regions with clear boundaries and uniform abrasion characteristics.

[0054] Step S34: Cluster the multiple candidate formulation adjustment region information to obtain multiple clusters.

[0055] In some embodiments, the K-means clustering algorithm can be used to cluster the information of the multiple candidate formulation adjustment regions to obtain multiple clusters.

[0056] K-means clustering is a typical unsupervised learning algorithm. It iteratively divides a dataset into K predefined clusters, minimizing the sum of squared distances from each sample point to the center of its cluster. In some embodiments, the value of K can be pre-set manually.

[0057] Multiple clusters are formed by dividing multiple candidate formulation adjustment regions according to the similarity of abrasion characteristics using the K-means clustering algorithm. The number of clusters is the K value in the K-means clustering algorithm.

[0058] Each of the multiple clusters represents a group of regions with similar abrasion characteristics or geometric properties. The similarity of regions within a cluster is high, while the differences between regions between clusters are significant.

[0059] As an example, specifically: K candidate formula adjustment regions are randomly selected as initial cluster centers. For each candidate formula adjustment region, the feature distance between the region and all initial cluster centers is calculated, and each candidate formula adjustment region is assigned to the nearest cluster center, thus forming K region clusters. For each cluster, the average value of the feature vectors of all regions within the cluster is calculated, and this average value is used as the new cluster center. The above assignment and update steps are repeated until the number of cluster centers reaches a preset number of iterations, at which point clustering is complete.

[0060] Using clustering algorithms to obtain K regional feature clusters, candidate formulation adjustment areas with similar abrasion characteristics and location attributes can be grouped into the same cluster, thereby achieving structured classification and management of dispersed areas. This operation can distinguish the core feature differences between different groups of candidate formulation adjustment areas, while reducing local feature fluctuations within clusters. It also allows the distribution pattern of adjustment areas on the surface of turbine flow components to be presented in clusters, which helps in subsequent targeted feature extraction and formulation optimization decisions for sets of candidate formulation adjustment areas with similar attributes.

[0061] Step S35: Based on the multiple clusters, determine the test formula adjustment area of ​​the turbine flow component and multiple other formula adjustment areas.

[0062] In some embodiments, a region selection model can be used to determine the test region for formula adjustment and multiple other regions for formula adjustment of the turbine flow components. The region selection model is a deep neural network. The input to the region selection model is the plurality of clusters, and the output of the region selection model is the test region for formula adjustment and multiple other regions for formula adjustment of the turbine flow components.

[0063] The test area for the water turbine flow components, which is the surface plate selected from multiple clusters by the region screening model, is the most representative and is used for actual coating abrasion testing.

[0064] The test areas of the turbine's flow-through components exhibit typical abrasion characteristics, reflecting the overall abrasion status of the components. These areas are the priority targets for abrasion testing in the coating formulation optimization process.

[0065] The remaining areas to be adjusted are a collection of other turbine surface plates that need to be optimized based on the test results, in addition to the areas to be adjusted for testing.

[0066] Multiple clusters illustrate the distribution characteristics of all regions to be adjusted. Each cluster's characteristics represent a region's behavior under abrasion conditions, providing the model with complete structured information for selecting the most representative test areas from multiple clusters and identifying surrounding areas requiring coordinated optimization. By analyzing the cluster size and central characteristics, the model can identify the regions that best cover the overall abrasion pattern.

[0067] The deep neural network evaluates multiple clusters of input data using a multi-criteria approach through its discriminative layer. The model analyzes the centroid features of each cluster and calculates the weight contribution of each cluster to the overall component erosion distribution. The deep neural network then selects the target cluster with the most prevalent erosion depth and the strongest representativeness of the operating conditions, directly designating this target cluster as the test area for coating adjustment in the turbine's flow components. After determining the test area, the model uses an attention mechanism to evaluate the correlation strength between the remaining clusters and the test area, determining whether the erosion evolution trend is consistent. The deep neural network traverses all clusters, excluding those with too slight erosion and no coating adjustment value, and then defines the remaining clusters with significant erosion characteristics as several other areas requiring coating adjustment. The model establishes an index relationship for each area to ensure that each area requiring adjustment corresponds to the relevant cluster features, thus completing the classification and characterization of the parts to be adjusted. Finally, the model determines the test area for coating adjustment and several other areas requiring coating adjustment.

[0068] In some embodiments, determining the test formula adjustment area and multiple remaining formula adjustment areas of the turbine flow components based on the plurality of clusters includes steps S351~S353:

[0069] Step S351: Determine the geometric morphological complexity level, abrasion severity level, and inter-cluster abrasion correlation matrix for each cluster based on the multiple clusters.

[0070] In some embodiments, deep neural networks can be used to determine the geometric complexity level, abrasion severity level, and inter-cluster abrasion correlation matrix for each cluster.

[0071] The geometrical complexity level of each cluster is a rating index used to measure the regularity of the shape of the turbine flow component area corresponding to a single cluster. The geometrical complexity level can be used to distinguish the regularity of the area's shape and the ease of spraying application.

[0072] The erosion severity level of each cluster is a grade index used to characterize the strength of surface erosion damage in the turbine flow component area corresponding to a single cluster.

[0073] The inter-cluster erosion correlation matrix is ​​a matrix data used to quantify the similarity of erosion characteristics and the consistency of evolutionary trends between any two clusters. The inter-cluster erosion correlation matrix reflects the degree of similarity and close correlation of erosion characteristics between different clusters.

[0074] The higher the value in the inter-cluster erosion correlation matrix, the more similar the erosion characteristics of the two clusters are, and the more consistent their working conditions and wear patterns are. In this case, a more similar coating formulation can be used for subsequent adaptation.

[0075] Deep neural networks can perform global perception and cluster-by-cluster analysis of the spatial range and surface state corresponding to multiple clusters. Deep neural networks can extract spatial information such as the boundary contour and shape regularity of each cluster, and then perform calculations through fully connected layers and nonlinear activation units to output the geometric complexity level of each cluster. Deep neural networks can also collect damage information such as scour intensity, cavitation degree, and sediment abrasion amount corresponding to each cluster, and output the erosion severity level of each cluster after comprehensive evaluation through a multilayer perceptron. Furthermore, deep neural networks can compare the similarity between any two clusters in terms of erosion mode, stress conditions, and working environment, and then generate an inter-cluster erosion correlation matrix through matrix mapping to fully present the correlation between different regions.

[0076] Step S352: Based on the geometric morphology complexity level, abrasion severity level, and abrasion correlation matrix between clusters, determine multiple candidate test areas, abrasion representativeness score, and process construction difficulty score for each candidate test area.

[0077] In some embodiments, a deep neural network can be used to determine multiple candidate test areas, an abrasion representativeness score for each candidate test area, and a process construction difficulty score.

[0078] Multiple candidate test regions are a set of turbine flow-through component regions that are typical of the working conditions and feasible for construction, selected from all clusters by a deep neural network.

[0079] Multiple candidate test regions are used to provide optional objects for the final test region determination.

[0080] The erosion representativeness score for each candidate test area is a numerical indicator used to quantify the typicality of the erosion conditions of a single candidate test area for the overall turbine flow components.

[0081] The abrasion representativeness score is used to measure the experimental reference value of the area.

[0082] The process construction difficulty score for each candidate test area is a numerical indicator used to evaluate the ease or difficulty of conducting coating spraying and abrasion testing on a single candidate test area.

[0083] The process and construction difficulty score is used to differentiate the on-site implementation feasibility of a region.

[0084] After receiving three types of input information—geometric complexity level, abrasion severity level, and inter-cluster abrasion correlation matrix—the deep neural network can perform feature fusion and logical judgment through a multilayer perceptron. Based on the abrasion severity level and the inter-cluster abrasion correlation matrix, the deep neural network can identify regions with strong representativeness of the working conditions, thus forming multiple candidate test regions. The model can perform weighted calculations by combining the matching degree between the abrasion distribution of the clusters and the global working condition distribution, and then output an abrasion representativeness score for each candidate test region. The model can also perform a comprehensive evaluation based on the geometric complexity level to determine the process construction difficulty score for each candidate test region.

[0085] Step S353: Based on the multiple candidate test areas, the erosion representativeness score of each candidate test area, and the process construction difficulty score, determine the test formula adjustment area of ​​the turbine flow component and multiple other formula adjustment areas.

[0086] In some embodiments, a deep neural network can be used to determine the test recipe adjustment area and multiple other recipe adjustment areas of the turbine flow components.

[0087] Deep neural networks can comprehensively analyze and select the best candidate test areas from multiple candidate test areas, considering both the abrasion representativeness score and the process construction difficulty score. They can weight and rank the abrasion representativeness score and process construction difficulty score, selecting the optimal area with strong abrasion representativeness and low process construction difficulty as the test area for adjusting the formula in the turbine's flow components. Furthermore, deep neural networks can classify all candidate test areas based on the abrasion representativeness score and process construction difficulty score, defining the effective areas other than the test area for formula adjustment as multiple remaining areas for formula adjustment, ultimately outputting the determined test area for formula adjustment and the multiple remaining areas for formula adjustment.

[0088] Step S4: Based on the abrasion distribution information of the test area to be adjusted, determine multiple identical test samples and multiple preliminary coating formulation information for the corresponding turbine flow components. The graphene content and the ratio of ceramic powder in each preliminary coating formulation information are different.

[0089] In some embodiments, a test specimen determination model can be used to determine multiple identical test specimens and multiple preliminary coating formulation information for a corresponding turbine flow component. The test specimen determination model is a Transformer model. The input to the test specimen determination model is the abrasion distribution information of the test area to be adjusted, and the output of the test specimen determination model is multiple identical test specimens and multiple preliminary coating formulation information for the corresponding turbine flow component.

[0090] The Transformer model is a deep learning model based on a self-attention mechanism. It abandons traditional recurrent or convolutional structures, directly capturing the global dependencies between any two elements in a sequence through multi-head self-attention. The Transformer model includes an encoder and a decoder. It excels at tasks with complex spatiotemporal relationships or parameter coupling, learning the deep logic between input features and the output target from mappings.

[0091] Multiple identical test specimens of the turbine flow-through components are standardized sets of metal entities fabricated based on the material, size, and surface morphology of the areas to be tested and adjusted according to the formulation, used for offline abrasion experiments. The consistent microstructure and surface characteristics of these multiple identical test specimens of the turbine flow-through components ensure comparability across different coating tests.

[0092] Multiple preliminary coating formulation information is a collection of graphene ceramic coating material design schemes with different component ratios, initially proposed based on the abrasion characteristics of the test area to be adjusted.

[0093] Preliminary coating formulation information includes graphene content and ceramic powder ratio.

[0094] The formulation information for ceramic powders includes the specific types of ceramic powders and the mass percentage of each type.

[0095] The graphene content and ceramic powder ratio differ in each preliminary coating formulation. Each preliminary coating formulation represents a potential direction for optimization.

[0096] Testing the abrasion distribution information in the area to be adjusted provides a precise basis for experimental design. The abrasion depth and environmental constraints in the abrasion distribution information can determine the damage intensity to be simulated for the test specimen, and at the same time establish the wear resistance performance benchmark that the initial coating formulation must meet. Testing the abrasion distribution information in the area to be adjusted provides targeted constraints for the preparation of test specimens and provides an environmental adaptation basis for coating formulation design, so as to ensure that the experimental design is accurately matched with actual working conditions.

[0097] The Transformer model leverages its powerful encoding capabilities to transform the abrasion distribution information of the test area to be adjusted into high-dimensional abrasion feature codes. Through a self-attention mechanism, the Transformer model analyzes every numerical point in the abrasion distribution information and identifies the pressure peak and particle impact frequency at the most severe abrasion locations. Based on learned material fatigue laws, the Transformer model first determines the number of comparative experiments needed at the output, thus determining the scale of multiple identical test samples for the turbine's flow components. Simultaneously, the decoder performs interactive deduction based on the high-dimensional abrasion feature codes, simulating the reinforcing effect of different proportions of graphene in the ceramic matrix, thereby determining the logic of graphene content's impact on toughness and hardness within a specific range. The Transformer model can generate multiple sets of significantly different parameter combinations, each representing a different attempt at graphene content and ceramic powder ratio. Through parallel processing via a multi-head mechanism, the Transformer model can balance cost, workability, and expected wear resistance, ultimately outputting multiple preliminary coating formulations containing detailed component ratios and matching them to corresponding sample requirements.

[0098] Step S5: Obtain erosion test data for the same test sample of the turbine flow-through component under each preliminary coating formulation information.

[0099] The abrasion test data for each preliminary coating formulation is obtained by performing abrasion tests on test specimens coated with different preliminary coating formulations using an abrasion testing machine. The abrasion test data for the same test specimens of the turbine flow components under each preliminary coating formulation includes the abrasion mass loss of the test specimen, the morphological change image of the surface coating, the change in the bonding strength between the coating and the substrate, the abrasion area, and the test duration.

[0100] Step S6: Based on the erosion test data of the same test sample of the turbine flow component under each preliminary coating formulation information, determine the target coating formulation information for the test area to be adjusted.

[0101] In some embodiments, a formulation adjustment determination model can be used to determine the target coating formulation information for the test area to be adjusted. The formulation adjustment determination model is a deep neural network. The input to the formulation adjustment determination model is the erosion test data of the same test sample of the turbine flow component under each preliminary coating formulation information, and the output of the formulation adjustment determination model is the target coating formulation information for the test area to be adjusted.

[0102] The target coating formulation information for the test area is the final formulation scheme of the graphene ceramic coating that is most suitable for the application, obtained by optimizing the formulation determination model for the test area.

[0103] The target coating formulation information for the area to be tested for formulation adjustment includes the optimal content of graphene and the optimal ratio of ceramic powder.

[0104] The erosion test data of identical test samples of the turbine's flow-through components, under each preliminary coating formulation, recorded the actual performance of each formulation scheme. The erosion test data is direct evidence for evaluating the quality of the formulation. This erosion test data allows the model to accurately pinpoint the extreme points of the performance curve by comparing the advantages and disadvantages of different preliminary coating formulations. Consequently, the model can accurately determine the target coating formulation for the testing area requiring formulation adjustment.

[0105] Deep neural networks can construct a performance evaluation subnetwork to normalize abrasion test data for identical test samples of turbine flow components under each initial coating formulation, and then calculate the comprehensive score of each formulation in terms of wear resistance, spalling resistance, and process economy. Utilizing the regression analysis capabilities in its hidden layers, the deep neural network can fit a nonlinear response surface between graphene content, ceramic powder ratio, and abrasion loss. The deep neural network analyzes which formulation combination achieves the minimum loss on this response surface, or within which parameter range performance improvement is most robust. Through continuous internal logic iteration, the deep neural network identifies redundant components or deficiencies in the initial formulation and performs weighted fusion and fine-tuning optimization on the best-performing formulations. Finally, the model outputs an optimal set of parameters that can achieve long-term protection under the current abrasion environment of the test area, thereby determining the target coating formulation information for the test area requiring formulation adjustment.

[0106] Step S7: Determine the target coating formula information for each of the remaining areas to be adjusted based on the target coating formula information of the tested area to be adjusted.

[0107] In some embodiments, Figure 4 This is a schematic flowchart illustrating a process for determining the target coating formulation information for each remaining region to be adjusted, as provided in an embodiment of the present invention. The determination of the target coating formulation information for each remaining region to be adjusted includes steps S71 to S73:

[0108] Step S71: Generate multiple simulated coating formula information for each of the remaining formula adjustment areas based on the target coating formula information of the test formula adjustment area.

[0109] In some embodiments, a formulation information simulation model can be used to generate multiple simulated coating formulation information for each of the remaining regions to be formulated and adjusted. The formulation information simulation model is a generative adversarial network (GAN). The input to the formulation information simulation model is the target coating formulation information for the tested region to be formulated and adjusted, and the output of the formulation information simulation model is multiple simulated coating formulation information for each of the remaining regions to be formulated and adjusted.

[0110] Generative Adversarial Networks (GANs) are deep learning models that learn data distributions by having two neural networks compete against each other. The generator aims to produce simulated data that matches the real data distribution, while the discriminator aims to distinguish between real data and the simulated data generated by the generator. Through continuous iterative training, the generator can produce highly realistic simulated data.

[0111] Multiple simulated coating formulation information for each of the remaining areas to be adjusted is a set of potential formulation alternatives for the remaining different abrasion areas, generated by the formulation information simulation model.

[0112] For each remaining region requiring formulation adjustment, the simulated coating formulation information includes the estimated graphene content and the estimated ceramic powder ratio to match the characteristics of the corresponding abrasive region.

[0113] The target coating formulation information for the testing area is the core basis for the entire coating formulation optimization process. The target coating formulation information, verified through abrasion testing, possesses excellent performance adapted to the abrasion environment of the testing area and provides a practically validated material composition basis for the formulation derivation of other areas requiring adjustment. While the coating formulations for other areas need to be adapted according to their own abrasion characteristics, they must adhere to the same graphene ceramic material system and component adaptation logic as the target coating formulation information. Therefore, the target coating formulation information, as the input basis for the formulation information simulation model, ensures that all candidate formulations generated by the model conform to the material properties and formulation design specifications of the graphene ceramic coating, thereby guaranteeing the rationality and feasibility of the formulation derivation.

[0114] The generator in the generative adversarial network (GAN) first takes the target coating formulation information of the test area as input. The generator learns the constraints and physical synergies between the components in the target formulation. Then, it introduces specific random perturbations based on the abrasion intensity difference coefficients between the remaining areas to be adjusted and the test area. The model attempts to adjust the graphene content fluctuations and the particle size distribution of the ceramic powder, thereby generating multiple formulation schemes with subtle differences for each of the remaining areas. Simultaneously, the discriminator verifies the formulations generated by the generator based on learned material principles and historical process constraints, eliminating schemes that are physically impossible to achieve or have extremely unreasonable component ratios. Through the cyclical adversarial interaction between the generator and the discriminator, the model continuously optimizes its generation strategy, ensuring that the generated formulations retain the best aspects of the target formulation while also adapting to the specific working conditions of the remaining areas. Finally, the model outputs multiple sets of simulated coating formulation information that have passed preliminary logical verification for each of the remaining areas to be adjusted.

[0115] Step S72: Construct a coating formulation simulation map. The coating formulation simulation map includes multiple nodes of other regions to be adjusted and edges between multiple other regions to be adjusted. The node features of each region to be adjusted include the abrasion distribution information of the other regions to be adjusted, multiple simulated coating formulation information of each other region to be adjusted, the abrasion distribution information of the test region to be adjusted, and the target coating formulation information of the test region to be adjusted.

[0116] The coating formulation simulation map consists of multiple nodes and multiple edges. The nodes represent multiple regions where the formulation needs adjustment, and an edge is established between every two of these regions. Each node and edge has corresponding features.

[0117] The remaining regions to be adjusted nodes represent the regions on the turbine flow components other than the test region that need to be adjusted. The node features of the remaining regions to be adjusted nodes include the abrasion distribution information of the remaining regions to be adjusted, the multiple simulated coating formula information of each remaining region to be adjusted, the abrasion distribution information of the test region to be adjusted, and the target coating formula information of the test region to be adjusted.

[0118] An edge represents the spatial relationship between multiple remaining regions to be adjusted in the recipe. The characteristics of an edge include the straight-line distance and spatial orientation between each remaining region to be adjusted in the recipe.

[0119] The abrasive environment and spatial location of the remaining regions requiring formulation adjustment directly affect the adaptability and uniformity of the coating formulation. By constructing a coating formulation simulation map, the spatial relationships and formulation correlation characteristics among the remaining regions requiring formulation adjustment can be clearly represented.

[0120] Step S73: Process the coating formulation simulation map based on the coating formulation optimization model to obtain the target coating formulation information.

[0121] The coating formulation optimization model is a graph neural network model. The input of the coating formulation optimization model is the coating formulation simulation map, and the output of the coating formulation optimization model is the target coating formulation information.

[0122] Graph Neural Network (GNN) models consist of Graph Neural Networks (GNNs) and fully connected layers. A GNN is a deep learning model that performs computations directly on graph data. Through message passing mechanisms, GNNs enable each node to aggregate feature information from its neighbors, thereby capturing complex structural information and long-range dependencies in the graph. GNNs include modules such as graph convolutional layers and graph attention layers, enabling deep mining of node embeddings through multi-layer iterative updates, thus achieving deep mining of node attributes, edge attributes, and overall graph features.

[0123] The target coating formulation information is a complete graphene ceramic coating formulation scheme that is finally determined through a coating formulation optimization model, covering all areas to be adjusted in the flow-through components of the turbine. The target coating formulation information includes the optimal graphene content and the optimal ratio of ceramic powder for each of the remaining areas to be adjusted.

[0124] The target coating formulation information enables the matching of globally optimized, proprietary coating ratios to each abrasion zone of the turbine's flow components. This target coating formulation information can be directly adapted for overall coating protection application.

[0125] Coating formulation simulation maps can integrate scattered formulation information from other regions requiring formulation adjustment into a structured system, eliminating the isolation of formulation data in a single region. Through the relationships between nodes and edges in the coating formulation simulation map, the distribution pattern and spatial relationships of erosion characteristics in each region requiring formulation adjustment can be systematically characterized. Coating formulation simulation maps provide a global analytical perspective for coating formulation optimization models, enabling the models to move beyond local formulation adaptation analysis in a single region and instead comprehensively consider the synergy and rationality of coating formulations across all regions.

[0126] The node features of the coating formulation simulation map include abrasion distribution information for each region to be adjusted, simulated coating formulation information, and abrasion distribution information and target coating formulation information for the tested region to be adjusted. These node features comprehensively reflect the abrasion damage degree, operating condition differences, and component characteristics of candidate formulations for each region to be adjusted. The abrasion distribution information provides operating condition basis for formula suitability judgment, while the simulated and target coating formulation information provide basic component references for model optimization. By analyzing these node features, the coating formulation optimization model can accurately identify the optimization space for each region to be adjusted in terms of formulation suitability and component rationality.

[0127] The edges of the coating formulation simulation map can effectively convey the spatial correlation information between the remaining regions to be adjusted. The straight-line distance and spatial orientation features contained in the edges reflect the spatial layout relationship of different regions to be adjusted on the turbine flow components. This feature is an important indicator for evaluating the overall consistency of the coating formulation system. In processing the coating formulation simulation map, the coating formulation optimization model can comprehensively consider the synergistic matching degree of the formulations in each region based on the edge features, and combine the adaptability of single-region formulations with the spatial correlation of the entire domain to ensure that the final output target coating formulation information not only meets the independent erosion protection requirements of each region, but also achieves the stability and uniformity of the overall coating system.

[0128] When processing the coating formulation simulation map, the graph neural network (Graph Neural Network) first initiates message passing, allowing each remaining node in the region to be adjusted to read the abrasion distribution information and multiple simulated formulation information from its node features. The Graph Neural Network calculates the edge weights between the node and its neighbors, with the weights determined by the physical distance and orientation between regions. Through multi-layer graph convolution operations, the Graph Neural Network diffuses the success parameters from the target coating formulation information of the test region along the edge structure to the entire graph, enabling each remaining node to perceive the performance of the baseline formulation. Subsequently, the Graph Neural Network uses an attention mechanism to filter and fuse multiple simulated coating formulation information within each node. The model can analyze which set of simulation parameters, combined with the abrasion pressure of neighboring nodes, produces the best protective expectation. The model continuously updates the latent features of each node while eliminating formulation conflicts caused by regional isolation, ensuring a smooth transition in the physical properties of coatings in adjacent regions. Finally, the Graph Neural Network summarizes and decodes all nodes in the entire coating formulation simulation map, then locks a unique formulation parameter for each region node in the coating formulation simulation map, thereby determining the target coating formulation information covering the entire machine.

[0129] Based on the same inventive concept Figure 5This invention provides a schematic diagram of a data analysis-based collaborative optimization system for graphene ceramic coating formulations in hydro-turbines. The system includes:

[0130] The acquisition module 81 is used to acquire historical flow field data and images of the turbine's flow passage components.

[0131] The erosion distribution information generation module 82 is used to generate erosion distribution information of the turbine flow-through component based on the historical flow field data of the turbine flow-through component and the image of the turbine flow-through component;

[0132] The adjustment area determination module 83 is used to determine the test formula adjustment area and multiple other formula adjustment areas of the turbine flow component based on the erosion distribution information of the turbine flow component and the image of the turbine flow component;

[0133] The preliminary formulation and sample determination module 84 is used to determine multiple identical test samples and multiple preliminary coating formulation information of the corresponding turbine flow components based on the abrasion distribution information of the test area to be adjusted. The graphene content and the ratio of ceramic powder in each preliminary coating formulation information are different.

[0134] The abrasion test data acquisition module 85 is used to acquire abrasion test data of the same test sample of the turbine flow-through component under each preliminary coating formulation information.

[0135] The test area target formula determination module 86 is used to determine the target coating formula information of the test area to be adjusted based on the erosion test data of the same test sample of the turbine flow component under each preliminary coating formula information.

[0136] The remaining area target formulation determination module 87 is used to determine the target coating formulation information for each remaining area to be adjusted based on the target coating formulation information of the test area to be adjusted.

[0137] 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.

[0138] 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 method for synergistic optimization of graphene ceramic coating formulations for hydro turbines based on data analysis, characterized in that, include: Acquire historical flow field data and images of the turbine's flow-through components; Based on the historical flow field data of the turbine flow passage components and the images of the turbine flow passage components, erosion distribution information of the turbine flow passage components is generated; Based on the erosion distribution information of the turbine flow-through components and the image of the turbine flow-through components, the test formula adjustment area and several other formula adjustment areas of the turbine flow-through components are determined. Based on the abrasion distribution information of the test area to be adjusted, multiple identical test samples and multiple preliminary coating formulation information of the corresponding turbine flow components were determined. The graphene content and the ratio of ceramic powder in each preliminary coating formulation information are different. Obtain erosion test data for identical test samples of the turbine flow components under each preliminary coating formulation. Based on the erosion test data of the same test sample of the turbine flow component under each preliminary coating formulation information, the target coating formulation information of the test area to be adjusted is determined; Based on the target coating formulation information of the test formulation adjustment area, the target coating formulation information for each of the remaining formulation adjustment areas is determined.

2. The method for collaborative optimization of graphene ceramic coating formulation for water turbines based on data analysis as described in claim 1, characterized in that, Based on the erosion distribution information of the turbine flow-through components and the image of the turbine flow-through components, the test formula adjustment area of ​​the turbine flow-through components and several other formula adjustment areas are determined, including: Based on the erosion distribution information of the turbine flow-through components and the image of the turbine flow-through components, multiple erosion contour lines of the turbine flow-through components are determined; Based on the erosion distribution information of the turbine flow-through components and the multiple erosion contour lines of the turbine flow-through components, the boundary point information of multiple candidate formula adjustment areas is determined; Based on the erosion distribution information of the turbine flow components and the boundary point information of the multiple candidate formula adjustment areas, multiple candidate formula adjustment area information is determined; Multiple clusters are obtained by clustering based on the adjustment region information of the multiple candidate formulas; Based on the multiple clusters, the test formula adjustment area for the turbine flow components and multiple other formula adjustment areas are determined.

3. The method for collaborative optimization of graphene ceramic coating formulation for water turbines based on data analysis as described in claim 1, characterized in that, The determination of target coating formula information for each remaining region to be adjusted based on the target coating formula information of the tested region includes: Based on the target coating formulation information of the test formulation adjustment area, generate multiple simulated coating formulation information for each of the remaining formulation adjustment areas; A coating formulation simulation map is constructed. The coating formulation simulation map includes multiple nodes of other regions to be adjusted and the edges between multiple other regions to be adjusted. The node features of each region to be adjusted include the abrasion distribution information of the other regions to be adjusted, multiple simulated coating formulation information of each other region to be adjusted, abrasion distribution information of the test region to be adjusted, and target coating formulation information of the test region to be adjusted. The target coating formulation information is obtained by processing the simulation map of the coating formulation based on the coating formulation optimization model.

4. The method for collaborative optimization of the formulation of graphene ceramic coating for water turbines based on data analysis as described in claim 3, characterized in that, The input to the coating formulation optimization model is the coating formulation simulation map, and the output of the coating formulation optimization model is the target coating formulation information.

5. A data-driven system for collaborative optimization of graphene ceramic coating formulations for hydro turbines, characterized in that... include: The acquisition module is used to acquire historical flow field data and images of the turbine's flow-through components. The erosion distribution information generation module is used to generate erosion distribution information of the turbine flow-through components based on the historical flow field data of the turbine flow-through components and the images of the turbine flow-through components; The adjustment area determination module is used to determine the test formula adjustment area and multiple other formula adjustment areas of the turbine flow components based on the erosion distribution information of the turbine flow components and the image of the turbine flow components; The preliminary formulation and sample determination module is used to determine multiple identical test samples and multiple preliminary coating formulation information of the corresponding turbine flow components based on the abrasion distribution information of the test area to be adjusted. The graphene content and the ratio of ceramic powder in each preliminary coating formulation information are different. The abrasion test data acquisition module is used to acquire abrasion test data of the same test sample of the turbine flow-through component under each preliminary coating formulation information; The test area target formulation determination module is used to determine the target coating formulation information of the test area to be adjusted based on the erosion test data of the same test sample of the turbine flow component under each preliminary coating formulation information. The remaining area target formulation determination module is used to determine the target coating formulation information for each remaining area to be adjusted based on the target coating formulation information of the test area to be adjusted.

6. The data analysis-based collaborative optimization system for water turbine graphene ceramic coating formulations as described in claim 5, characterized in that, The adjustment area determination module is also used for: Based on the erosion distribution information of the turbine flow-through components and the image of the turbine flow-through components, multiple erosion contour lines of the turbine flow-through components are determined; Based on the erosion distribution information of the turbine flow-through components and the multiple erosion contour lines of the turbine flow-through components, the boundary point information of multiple candidate formula adjustment areas is determined; Based on the erosion distribution information of the turbine flow components and the boundary point information of the multiple candidate formula adjustment areas, multiple candidate formula adjustment area information is determined; Multiple clusters are obtained by clustering based on the adjustment region information of the multiple candidate formulas; Based on the multiple clusters, the test formula adjustment area for the turbine flow components and multiple other formula adjustment areas are determined.

7. The data analysis-based collaborative optimization system for graphene ceramic coating formulations in water turbines as described in claim 5, characterized in that, The remaining region target formula determination module is also used for: Based on the target coating formulation information of the test formulation adjustment area, generate multiple simulated coating formulation information for each of the remaining formulation adjustment areas; A coating formulation simulation map is constructed. The coating formulation simulation map includes multiple nodes of other regions to be adjusted and the edges between multiple other regions to be adjusted. The node features of each region to be adjusted include the abrasion distribution information of the other regions to be adjusted, multiple simulated coating formulation information of each other region to be adjusted, abrasion distribution information of the test region to be adjusted, and target coating formulation information of the test region to be adjusted. The target coating formulation information is obtained by processing the simulation map of the coating formulation based on the coating formulation optimization model.

8. The data analysis-based collaborative optimization system for water turbine graphene ceramic coating formulations as described in claim 5, characterized in that, The input to the coating formulation optimization model is the coating formulation simulation map, and the output of the coating formulation optimization model is the target coating formulation information.

9. An electronic device, characterized in that, include: processor; 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 data analysis-based co-optimization method for hydro-turbine graphene ceramic coating formulations as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the data analysis-based collaborative optimization method for the formulation of graphene ceramic coatings for water turbines as described in any one of claims 1 to 4.

Images