Method and system for predicting corrosion resistance of zinc-aluminum-magnesium coating

Abstract

The invention relates to the technical field of coating detection, and discloses a zinc-aluminum-magnesium coating corrosion resistance prediction method and system.The method comprises the steps that a coating sample microstructure image is segmented, a substance area is recognized, and structure partition data is obtained; carrying out multi-dimensional coupling on the geometrical morphology and the spatial distribution characteristics to obtain a comprehensive morphology characteristic set; fusing the chemical component data and the comprehensive morphology feature set, and performing feature dimension reduction processing on the fused feature set to obtain a standard feature identifier; constructing a corrosion resistance knowledge graph by taking the standard feature identifier as a node and taking an incidence relation of actually measured corrosion resistance grades in the zinc-aluminum-magnesium coating sample as an edge; inputting original data of a plating layer to be predicted, and identifying similar sample nodes; connecting the similar sample nodes into a relation path of the zinc-aluminum-magnesium coating to be predicted, and performing traversal exploration along the relation path to obtain a corrosion resistance prediction grade corresponding to the zinc-aluminum-magnesium coating to be predicted; according to the invention, the corrosion resistance prediction efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coating detection, in particular to a zinc-aluminum-magnesium coating corrosion resistance prediction method and system. BACKGROUND

[0002] Zinc-aluminum-magnesium coating is widely used in the industrial field due to its excellent protective performance, and its corrosion resistance is the core indicator for evaluating material quality and service life. Current methods for predicting the corrosion resistance of zinc-aluminum-magnesium coating rely on traditional experimental detection methods, which not only consume a large number of samples and time, but also can only be used for post-evaluation after the coating is prepared, and cannot guide the optimization of production process in advance. At the same time, these methods often only focus on single-dimensional features or isolated analysis of microstructure, chemical composition and other factors, ignoring the internal correlation between multiple features, resulting in a large deviation between the prediction results and the actual corrosion resistance, and it is difficult to meet the precise evaluation demand.

[0003] With the increasing demand for efficiency and accuracy in industrial production, some prediction methods attempt to integrate multi-dimensional data, but there are obvious deficiencies in feature processing: either the high-dimensional features after fusion are not effectively reduced, resulting in data redundancy and a sharp increase in computing cost; or there is a lack of stable data correlation model, and when facing coating samples under different process parameters, the prediction adaptability is poor and the result stability is insufficient, which cannot efficiently support rapid performance judgment in batch production, further increasing the research and production cost of enterprises, therefore how to improve the efficiency of zinc-aluminum-magnesium coating corrosion resistance prediction has become a problem to be solved. SUMMARY

[0004] The present application provides a zinc-aluminum-magnesium coating corrosion resistance prediction method and system to solve the problems raised in the background.

[0005] To achieve the above purpose, the present application provides a zinc-aluminum-magnesium coating corrosion resistance prediction method, which comprises:

[0006] S1, dividing the microstructure image of the zinc-aluminum-magnesium coating sample into different analysis levels, and identifying different material region data on the analysis level to obtain structure partition data of the zinc-aluminum-magnesium coating sample;

[0007] S2, multi-dimensionally coupling the geometric shape features and spatial distribution features in the structure partition data to obtain a comprehensive morphology feature set of the zinc-aluminum-magnesium coating sample;

[0008] S3, fusing the chemical composition data of the zinc-aluminum-magnesium coating sample and the comprehensive morphology feature set, and performing feature dimension reduction processing on the fused feature set to obtain a standard feature identifier of the zinc-aluminum-magnesium coating sample;

[0009] S4, constructing an anti-corrosion knowledge graph of the zinc-aluminum-magnesium plated sample by taking the standard feature label as a node and the correlation between the measured anti-corrosion grade of the zinc-aluminum-magnesium plated sample as an edge;

[0010] S5, inputting original analysis data of a zinc-aluminum-magnesium plated layer to be predicted into the anti-corrosion knowledge graph, and identifying a similar sample node of the zinc-aluminum-magnesium plated layer to be predicted;

[0011] S6, connecting the similar sample nodes into a relationship path of the zinc-aluminum-magnesium plated layer to be predicted, and traversing and exploring along the relationship path to obtain an anti-corrosion prediction grade corresponding to the zinc-aluminum-magnesium plated layer to be predicted.

[0012] In a preferred embodiment, the microstructure image of the zinc-aluminum-magnesium plated sample is divided into different analysis levels, and different material region data on the analysis levels are identified to obtain structure partition data of the zinc-aluminum-magnesium plated sample, including:

[0013] A microstructure image of a zinc-aluminum-magnesium plated sample is collected, and a multi-scale Gaussian filtering process is performed on the microstructure image to obtain a multi-scale image set of the zinc-aluminum-magnesium plated sample;

[0014] Boundary feature recognition is performed on the filtered regions between phases in the multi-scale image set to obtain phase boundary distribution information of the zinc-aluminum-magnesium plated sample;

[0015] According to the phase boundary distribution information, a region growing technique is used to segment the multi-scale image set to obtain independent partitions of the aluminum-magnesium plated sample;

[0016] Morphological optimization processing is performed on the independent partitions to obtain material regions of the zinc-aluminum-magnesium plated sample;

[0017] The geometric features and spatial relationship features in the material regions are quantified to obtain structure partition data of the zinc-aluminum-magnesium plated sample.

[0018] In a preferred embodiment, the geometric and spatial distribution features in the structure partition data are multi-dimensionally coupled to obtain a comprehensive topographic feature set of the zinc-aluminum-magnesium plated sample, including:

[0019] The region contour information in the structure partition data is analyzed to obtain geometric descriptors of the zinc-aluminum-magnesium plated sample;

[0020] According to the region spatial coordinates and spatial topological structure in the structure partition data, a region relationship network of the zinc-aluminum-magnesium plated sample is constructed;

[0021] The region distribution pattern in the region relationship network is identified to obtain a spatial distribution feature descriptor of the zinc-aluminum-magnesium plated sample;

[0022] The geometric descriptor and the spatial distribution feature descriptor are standardized and integrated to obtain the geometric morphology feature set and spatial distribution feature set of the zinc-aluminum-magnesium coating sample;

[0023] Tensor synthesis is performed on the geometric feature set and the spatial distribution feature set to obtain the geometric feature vector and spatial feature vector of the zinc-aluminum-magnesium coating sample;

[0024] The geometric feature vector and the spatial feature vector are coupled to obtain the comprehensive feature value of the zinc-aluminum-magnesium coating sample. The comprehensive feature value is then integrated to obtain the comprehensive morphological feature set of the zinc-aluminum-magnesium coating sample.

[0025] In a preferred embodiment, the formula for calculating the comprehensive feature value is:

[0026] ;

[0027] in, This represents the comprehensive feature value. Indicates the first The geometric feature vectors This represents the mean of the geometric feature vectors. Indicates the first The aforementioned spatial feature vectors The mean of the spatial feature vectors is represented by . This represents the preset weighting coefficient. This represents the number of all feature vectors.

[0028] In a preferred embodiment, the process of fusing the chemical composition data of the zinc-aluminum-magnesium coating sample and the comprehensive morphological feature set, and then performing feature dimensionality reduction on the fused feature set to obtain the standard feature identifier of the zinc-aluminum-magnesium coating sample, includes:

[0029] The chemical composition data of the zinc-aluminum-magnesium coating sample were normalized to obtain the standard composition set of the zinc-aluminum-magnesium coating sample.

[0030] The standard component set and the comprehensive morphological feature set are deeply coupled to obtain the joint feature set of the zinc-aluminum-magnesium coating sample;

[0031] Screen the feature subsets in the joint feature set that have a high correlation with the corrosion resistance level of the zinc-aluminum-magnesium coating samples;

[0032] Spatial compression is performed on the feature subset to obtain the standard feature identifier of the zinc-aluminum-magnesium coating sample.

[0033] In a preferred embodiment, the corrosion resistance knowledge graph of the zinc-aluminum-magnesium plated sample is constructed by taking the standard feature identifier as a node and the correlation between the measured corrosion resistance grade in the zinc-aluminum-magnesium plated sample as an edge, and comprises:

[0034] The standard feature identifier is subjected to spatial clustering analysis to obtain a feature similarity grouping of the zinc-aluminum-magnesium plated sample;

[0035] The standard feature identifier is mapped to a node, and a topological connection relationship of the node is established according to the feature similarity grouping to obtain an initial graph structure of the zinc-aluminum-magnesium plated sample;

[0036] The edges of the initial graph structure are subjected to weight amplitude according to the correlation strength of the measured corrosion resistance grade in the zinc-aluminum-magnesium plated sample to obtain an intermediate graph structure of the zinc-aluminum-magnesium plated sample;

[0037] The intermediate graph structure is subjected to graph structure optimization to eliminate isolated nodes in the intermediate graph structure to obtain the corrosion resistance knowledge graph of the zinc-aluminum-magnesium plated sample.

[0038] In a preferred embodiment, the original analysis data of the zinc-aluminum-magnesium plated layer to be predicted is input into the corrosion resistance knowledge graph, and similar sample nodes of the zinc-aluminum-magnesium plated layer to be predicted are identified, comprising:

[0039] The original analysis data of the zinc-aluminum-magnesium plated layer to be predicted is mapped to a feature space consistent with the zinc-aluminum-magnesium plated sample to obtain a temporary feature identifier of the zinc-aluminum-magnesium plated layer to be predicted;

[0040] The temporary feature identifier is subjected to multi-dimensional similarity comparison with the standard feature identifier in the corrosion resistance knowledge graph to obtain a candidate similar node of the zinc-aluminum-magnesium plated layer to be predicted;

[0041] Based on the topological structure of the corrosion resistance knowledge graph, the candidate similar node is subjected to domain expansion to obtain a correlation node of the zinc-aluminum-magnesium plated layer to be predicted;

[0042] The confidence of the correlation node is evaluated, and the node with high confidence is taken as a similar sample node of the zinc-aluminum-magnesium plated layer to be predicted.

[0043] In a preferred embodiment, the similar sample nodes are connected into a relationship path of the zinc-aluminum-magnesium plated layer to be predicted, and the relationship path is traversed and explored to obtain a corresponding corrosion resistance prediction grade of the zinc-aluminum-magnesium plated layer to be predicted, comprising:

[0044] According to the spatial distribution of the similar sample nodes in the corrosion resistance knowledge graph, an initial path framework of the zinc-aluminum-magnesium plated layer to be predicted is established;

[0045] Based on the correlation degree between the similarity of the similar sample node and the corrosion resistance level in the corrosion resistance knowledge graph, a priority evaluation is performed on the path in the initial path framework, to obtain a priority path framework of the zinc-aluminum-magnesium coating to be predicted;

[0046] A multi-path search strategy is adopted to explore the priority path framework, to obtain a candidate path of the zinc-aluminum-magnesium coating to be predicted;

[0047] The coherence of the candidate path is verified, to obtain a relationship path of the zinc-aluminum-magnesium coating to be predicted.

[0048] In a preferred embodiment, the connecting of the similar sample nodes into the relationship path of the zinc-aluminum-magnesium coating to be predicted, and the traversal exploration along the relationship path, to obtain the corrosion resistance prediction level corresponding to the zinc-aluminum-magnesium coating to be predicted, comprises:

[0049] Starting from the starting node of the relationship path, the similar sample nodes are traversed in priority according to the path weight of the relationship path, to obtain the corrosion resistance level characteristics of the similar sample nodes;

[0050] The corrosion resistance level characteristics are integrated, and the feature conflicts between the similar sample nodes are eliminated, to obtain a corrosion resistance level characteristic set of the zinc-aluminum-magnesium coating to be predicted;

[0051] The corrosion resistance level characteristic set is analyzed in terms of trends, to obtain the corrosion resistance prediction level of the zinc-aluminum-magnesium coating to be predicted.

[0052] To solve the above problems, the application further provides a zinc-aluminum-magnesium coating corrosion resistance prediction system, which comprises:

[0053] A data partition module is configured to divide a microstructure image of a zinc-aluminum-magnesium coating sample into different analysis levels, and identify different material region data on the analysis levels, to obtain structure partition data of the zinc-aluminum-magnesium coating sample;

[0054] A feature coupling module is configured to perform multi-dimensional coupling on geometric shape features and spatial distribution features in the structure partition data, to obtain a comprehensive topographic feature set of the zinc-aluminum-magnesium coating sample;

[0055] A standard feature identification generation module is configured to fuse chemical composition data of the zinc-aluminum-magnesium coating sample and the comprehensive topographic feature set, and perform feature dimension reduction processing on the fused feature set, to obtain a standard feature identification of the zinc-aluminum-magnesium coating sample;

[0056] An anti-corrosion property mapping module is configured to construct an anti-corrosion property knowledge graph of the zinc-aluminum-magnesium coating sample by taking the standard feature identifiers as nodes and the correlation between the measured anti-corrosion property grades of the zinc-aluminum-magnesium coating sample as edges.

[0057] A similar node identification module is configured to input the original analysis data of a zinc-aluminum-magnesium coating to be predicted into the anti-corrosion property knowledge graph and identify similar sample nodes of the zinc-aluminum-magnesium coating to be predicted.

[0058] An anti-corrosion property grade prediction module is configured to connect the similar sample nodes into a relationship path of the zinc-aluminum-magnesium coating to be predicted and explore along the relationship path to obtain an anti-corrosion property prediction grade corresponding to the zinc-aluminum-magnesium coating to be predicted.

[0059] Compared with the prior art, the present application has the following beneficial effects:

[0060] 1. The present application precisely obtains structure partition data by segmenting the microstructure image of the zinc-aluminum-magnesium coating sample into different analysis levels and identifying material region data; multi-dimensionally couples geometric morphological features and spatial distribution features to form a comprehensive morphological feature set, simultaneously fuses chemical composition data and performs feature dimension reduction processing to obtain standard feature identifiers. This series of operations comprehensively extracts and optimizes multi-dimensional key features of the coating, makes the feature characterization more consistent with the actual influencing factors of the coating anti-corrosion property, significantly improves the precision and effectiveness of the feature data, and provides a high-quality data basis for subsequent anti-corrosion property prediction.

[0061] 2. The present application constructs an anti-corrosion property knowledge graph by taking the standard feature identifiers as nodes and the correlation between the measured anti-corrosion property grades of the sample as edges; subsequently inputs the original analysis data of the coating to be predicted into the graph, precisely identifies similar sample nodes and connects them into a relationship path, and explores along the path to obtain a prediction grade. This process establishes a stable correlation model between features and anti-corrosion property, efficiently matches similar samples of the coating to be predicted, realizes precise deduction of the anti-corrosion property grade, effectively improves the anti-corrosion property prediction efficiency, and provides reliable technical support for performance evaluation and application selection of zinc-aluminum-magnesium coating materials. BRIEF DESCRIPTION OF DRAWINGS

[0062] Figure 1 A flowchart of a zinc-aluminum-magnesium coating anti-corrosion property prediction method provided by an embodiment of the present application is shown in the figure.

[0063] Figure 2 A functional module diagram of a zinc-aluminum-magnesium coating anti-corrosion property prediction system provided by an embodiment of the present application is shown in the figure.

[0064] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0065] It should be understood that the specific embodiments described herein are merely exemplary and not intended to limit the application.

[0066] The embodiment of the present application provides a zinc-aluminum-magnesium coating corrosion resistance prediction method. The execution subject of the zinc-aluminum-magnesium coating corrosion resistance prediction method includes but is not limited to at least one of electronic devices capable of being configured to execute the method provided by the embodiment of the present application, such as a server and a terminal. In other words, the zinc-aluminum-magnesium coating corrosion resistance prediction method can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

[0067] Referring to Figure 1 Fig. 1 is a flowchart of a zinc-aluminum-magnesium coating corrosion resistance prediction method provided by an embodiment of the present application. In the embodiment, the zinc-aluminum-magnesium coating corrosion resistance prediction method includes the following steps.

[0068] S1, dividing a microstructure image of a zinc-aluminum-magnesium coating sample into different analysis levels, and identifying different material region data on the analysis levels to obtain structure partition data of the zinc-aluminum-magnesium coating sample;

[0069] In the embodiment of the present application, the step of dividing a microstructure image of a zinc-aluminum-magnesium coating sample into different analysis levels, and identifying different material region data on the analysis levels to obtain structure partition data of the zinc-aluminum-magnesium coating sample includes the following steps.

[0070] Collecting a microstructure image of a zinc-aluminum-magnesium coating sample, and performing multi-scale Gaussian filtering processing on the microstructure image to obtain a multi-scale image set of the zinc-aluminum-magnesium coating sample;

[0071] Performing boundary feature identification on a filtering region between phases in the multi-scale image set to obtain phase boundary distribution information of the zinc-aluminum-magnesium coating sample;

[0072] According to the phase boundary distribution information, the multi-scale image set is segmented by using a region growing technique to obtain an independent partition of the aluminum-magnesium coating sample;

[0073] Performing morphological optimization processing on the independent partition to obtain a material region of the zinc-aluminum-magnesium coating sample;

[0074] Quantify the geometric features and spatial relationship features in the material region to obtain the structure partition data of the zinc-aluminum-magnesium coating sample.

[0075] When collecting the microstructure image of the zinc-aluminum-magnesium coating sample, a high-resolution microscopic imaging device is used to take pictures at different positions and different focal lengths of the coating sample, so that the image can clearly present the phase distribution, micro-morphology and various detail features inside the coating, and the complete original microstructure image is obtained. When performing multi-scale Gaussian filtering processing on the microstructure image, a plurality of different filtering scales are set, and the original image is smoothed for each scale. By weakening the random noise interference in the image, while completely retaining the key detail information of the phase under the corresponding scale, a corresponding filtered image is obtained after processing each scale. All filtered images of different scales are integrated to form a multi-scale image set of the zinc-aluminum-magnesium coating sample.

[0076] The filtering region in the multi-scale image set is a transition region formed after filtering different phases. When performing boundary feature recognition on this region, the filtering region in each scale image is analyzed one by one, and the gray value change, texture difference and contour direction of the filtering region and the adjacent phase region are observed. By capturing the continuous points of gray value mutation, the boundary lines between different phases are determined, and the phase type corresponding to each boundary is labeled. The boundary information identified in all scale images is summarized and integrated to finally obtain the phase boundary distribution information of the zinc-aluminum-magnesium coating sample.

[0077] According to the obtained phase boundary distribution information, the region growing technique is used to segment the multi-scale image set, taking the phase boundary labeled in the phase boundary distribution information as a clear constraint. In each scale image, a core pixel point inside the phase is selected as a growing starting point. According to the similarity standard of the gray value and texture features of the pixel point and the core pixel point, the growing is gradually expanded to the surrounding adjacent pixel points. The growing process strictly does not cross the identified phase boundary. When there is no pixel point that meets the similarity standard in the growing region, the growing is stopped. Each complete region formed by growing is an independent partition. After the segmentation operation of all scale images is completed, the independent partition of the zinc-aluminum-magnesium coating sample is obtained.

[0078] When performing morphological optimization processing on the independent partition, the internal defects of each independent partition are first processed, and the small holes formed by noise residues in the partition are filled, so that the internal structure of the partition is complete and has no missing part. Then, the edge problem is processed to eliminate the redundant structures such as burrs and irregular protrusions on the edge of the partition. By smoothing the edge pixel points, the edge lines are continuous and regular, while the overall contour and original size of the partition are not changed. Each independent partition after the above processing is the material region of the zinc-aluminum-magnesium coating sample.

[0079] When quantifying the geometric and spatial characteristics of material regions, for geometric characteristics, quantifiable indicators such as the area, perimeter, and regularity of the shape of each material region are statistically analyzed, and the specific values ​​of each indicator are directly recorded. For spatial characteristics, the specific position coordinates of each material region in the image are analyzed, the straight-line distance between different material regions is measured, and whether there is overlap and the extent of overlap are determined. At the same time, the arrangement and distribution of each material region are recorded. The quantification results of the geometric characteristics and spatial characteristics of all material regions are systematically integrated and summarized to finally obtain the structural partitioning data of the zinc-aluminum-magnesium coating sample.

[0080] The beneficial effect is that through a systematic and meticulous image processing and feature quantization process, the structural partition data of zinc-aluminum-magnesium coating samples were accurately obtained. This not only effectively eliminated image noise interference and clarified the accurate range of phase boundaries and material regions, but also comprehensively captured the geometric features and spatial relationship features of the material regions. This provides accurate and reliable basic data support for subsequent multi-dimensional feature coupling and corrosion resistance prediction, ensuring the scientific nature and accuracy of the entire prediction process.

[0081] S2. Perform multidimensional coupling on the geometric morphological features and spatial distribution features in the structural partition data to obtain the comprehensive morphological feature set of the zinc-aluminum-magnesium coating sample;

[0082] In this embodiment of the invention, the multidimensional coupling of geometric morphological features and spatial distribution features in the structural partitioning data to obtain the comprehensive morphological feature set of the zinc-aluminum-magnesium coating sample includes:

[0083] The geometric descriptor of the zinc-aluminum-magnesium coating sample is obtained by parsing the region contour information in the structural partition data.

[0084] Based on the regional spatial coordinates and spatial topology in the structural partitioning data, a regional relationship network of the zinc-aluminum-magnesium coating sample is constructed.

[0085] Identify the regional distribution patterns in the regional relationship network to obtain the spatial distribution feature descriptor of the zinc-aluminum-magnesium coating sample;

[0086] The geometric descriptor and the spatial distribution feature descriptor are standardized and integrated to obtain the geometric morphology feature set and spatial distribution feature set of the zinc-aluminum-magnesium coating sample;

[0087] Tensor synthesis is performed on the geometric feature set and the spatial distribution feature set to obtain the geometric feature vector and spatial feature vector of the zinc-aluminum-magnesium coating sample;

[0088] The geometric feature vector and the spatial feature vector are coupled to obtain a comprehensive feature value of the zinc-aluminum-magnesium coating sample, and the comprehensive feature value is integrated to obtain a comprehensive morphology feature set of the zinc-aluminum-magnesium coating sample.

[0089] The calculation formula of the comprehensive feature value is:

[0090] ;

[0091] wherein, represents the comprehensive feature value, represents the geometric feature vector, represents the geometric feature vector, represents the spatial feature vector, represents the spatial feature vector, represents the spatial feature vector, represents the spatial feature vector, represents a preset weight coefficient, represents the number of all feature vectors.

[0092] When analyzing the area contour information in the structure partition data, edge tracking is performed on each material area contained in the structure partition data one by one, the position information of the boundary pixel points of the material area is recorded in sequence, a complete and continuous area contour line is formed, key features are extracted based on the contour line, including the bending degree of the contour, the number of turns, the regularity of the overall shape, and the area range feature surrounded by the contour, and various information about the area contour is systematically arranged to obtain the geometric descriptor of the zinc-aluminum-magnesium coating sample.

[0093] When constructing the area relationship network according to the area spatial coordinates and the spatial topological structure in the structure partition data, the specific spatial coordinates of each material area are extracted from the structure partition data to determine the position of each area in the overall coating structure, and the spatial topological relationship between the material areas is analyzed, including adjacent contact relationship, containing relationship and separation relationship, etc. Each material area is regarded as an independent node in the network, the nodes with topological relationship are connected by lines, and the corresponding topological relationship type is marked on the line, and finally the area relationship network of the zinc-aluminum-magnesium coating sample is formed.

[0094] In identifying the area distribution pattern in the area relationship network, the arrangement state of all nodes in the area relationship network is observed as a whole, it is judged whether the nodes are uniformly distributed, concentratedly distributed or directionally arranged in a specific shape, the number and proportion of node pairs in different topological relationships in the network are counted, the dense degree distribution of the connection between nodes is analyzed, and whether there is a repeatedly appearing node association combination form is observed. The observation results and statistical information about the area arrangement mode and association characteristics are integrated to obtain the spatial distribution characteristic descriptor of the zinc-aluminum-magnesium plated layer sample.

[0095] In standardizing and integrating the geometric descriptors and the spatial distribution characteristic descriptors, the feature expression formats of the two are unified, the different types of contour features in the geometric descriptors are classified and arranged according to the preset unified classification rule, the arrangement mode and association proportion in the spatial distribution characteristic descriptors are also classified according to the corresponding rule, so that the expression modes of all features are consistent and unambiguous. Then, all the classified and arranged geometric descriptor related information is summarized into a book to obtain the geometric shape feature set of the zinc-aluminum-magnesium plated layer sample, and all the classified and arranged spatial distribution characteristic descriptor related information is summarized into a book to obtain the spatial distribution feature set of the zinc-aluminum-magnesium plated layer sample.

[0096] In tensor synthesis of the geometric shape feature set and the spatial distribution feature set, for the geometric shape feature set, each geometric feature is assigned a dedicated dimension position according to the attribute category order of the geometric features in the feature set, all the geometric features are arranged in order according to the assigned dimension positions to form an ordered feature combination, and a geometric feature vector of the zinc-aluminum-magnesium plated layer sample is obtained. For the spatial distribution feature set, the same way is adopted, dimension positions are assigned according to the attribute category order of the spatial features in the feature set, all the spatial features are arranged in order according to the dimension positions to form an ordered feature combination, and a spatial feature vector of the zinc-aluminum-magnesium plated layer sample is obtained.

[0097] The mean value of the geometric feature vector is calculated by adding up the feature values corresponding to all geometric feature vectors, and then dividing the total sum obtained by the total number of geometric feature vectors.

[0098] The mean value of the spatial feature vector is calculated by adding up the feature values corresponding to all spatial feature vectors, and then dividing the total sum obtained by the total number of spatial feature vectors.

[0099] The preset weight coefficient is a fixed value determined by the preliminary study and actual application verification of the influence law of the plated layer performance according to the importance of the geometric features and the spatial features to the corrosion resistance of the zinc-aluminum-magnesium plated layer.

[0100] The number of all feature vectors refers to the total number of geometric feature vectors obtained after tensor synthesis, and since the geometric feature vectors and the spatial feature vectors are generated based on the same batch of structural partition data processing, the number of the two is completely consistent.

[0101] The first geometric feature vector is the first vector in the geometric feature vectors obtained after tensor synthesis. The first geometric feature vector is the first vector in the geometric feature vectors obtained after tensor synthesis. The first geometric feature vector is the first vector in the geometric feature vectors obtained after tensor synthesis. The first geometric feature vector is the first vector in the geometric feature vectors obtained after tensor synthesis. The first geometric feature vector is the first vector in the geometric feature vectors obtained after tensor synthesis.

[0102] When the geometric feature vectors and the spatial feature vectors are coupled, the fusion of the two types of vectors is realized according to a specific calculation logic. During calculation, the difference between each geometric feature vector and the geometric feature vector mean value is first calculated, and the difference between each spatial feature vector and the spatial feature vector mean value is also calculated. Then, the two difference values at each corresponding position are multiplied, and then multiplied by a preset weight coefficient. The accumulated product results obtained are added to obtain the numerator value.

[0103] The difference between each geometric feature vector and the mean value is squared and accumulated, and the accumulated result is square rooted to obtain the first denominator factor. The difference between each spatial feature vector and the mean value is squared and accumulated, and the accumulated result is square rooted to obtain the second denominator factor. The two denominator factors are multiplied to obtain the denominator value.

[0104] Finally, the numerator value is divided by the denominator value to obtain the result, which is the comprehensive feature value. The core of this calculation process is to quantitatively reflect the correlation degree and comprehensive effect of the geometric feature vectors and the spatial feature vectors. Each comprehensive feature value can comprehensively reflect the synergistic effect of the geometric features and the spatial features in the corresponding dimension. The comprehensive feature values corresponding to all dimensions are arranged and summarized in the original dimension order to obtain the comprehensive morphology feature set of the zinc-aluminum-magnesium coating sample.

[0105] The beneficial effect is that through a series of consecutive and detailed feature extraction, integration, synthesis and coupling operations, the geometric shape and spatial distribution related features of the zinc-aluminum-magnesium coating sample are comprehensively and accurately captured, forming a complete comprehensive morphology feature set. This feature set not only retains the individual morphology characteristics of each material region of the coating, but also reflects the correlation rules between the regions, providing a comprehensive, reliable and strongly correlated morphology feature basis for subsequent fusion of chemical composition data for corrosion resistance prediction, ensuring the scientificity and accuracy of the subsequent prediction process.

[0106] S3, fuse the chemical composition data of the zinc-aluminum-magnesium plated layer sample and the comprehensive appearance feature set, and perform feature dimension reduction processing on the fused feature set to obtain a standard feature identifier of the zinc-aluminum-magnesium plated layer sample;

[0107] In the embodiment of the application, the fusion of the chemical composition data of the zinc-aluminum-magnesium plated layer sample and the comprehensive appearance feature set, and the feature dimension reduction processing on the fused feature set to obtain a standard feature identifier of the zinc-aluminum-magnesium plated layer sample, comprises:

[0108] The chemical composition data of the zinc-aluminum-magnesium plated layer sample is normalized to obtain a standard component set of the zinc-aluminum-magnesium plated layer sample;

[0109] The standard component set and the comprehensive appearance feature set are deeply coupled to obtain a joint feature set of the zinc-aluminum-magnesium plated layer sample;

[0110] A feature subset with a high correlation degree to the corrosion resistance grade of the zinc-aluminum-magnesium plated layer sample is screened from the joint feature set;

[0111] The feature subset is spatially compressed to obtain a standard feature identifier of the zinc-aluminum-magnesium plated layer sample.

[0112] When the chemical composition data of the zinc-aluminum-magnesium plated layer sample is normalized, first, the types and specific content data of the chemical elements contained in the sample are determined, the value range of the content of each chemical element is determined, the maximum value and the minimum value of the element in all detection data are found, the specific content of the element in each sample is subtracted from the minimum value, and the obtained difference is divided by the difference between the maximum value and the minimum value. Through this operation, the content data of all chemical elements is uniformly mapped to the same numerical interval, eliminating the dimensional differences between different element contents. After processing, all the chemical element content data is systematically arranged to obtain a standard component set of the zinc-aluminum-magnesium plated layer sample.

[0113] When the standard component set and the comprehensive appearance feature set are deeply coupled, a single zinc-aluminum-magnesium plated layer sample is taken as a basic unit, the normalized content data of all chemical elements in the standard component set of the sample is extracted, and all comprehensive feature values in the comprehensive appearance feature set of the sample are extracted. According to the corresponding logic of "chemical element content-comprehensive feature value", the two types of feature data of the same sample are matched and associated one by one to ensure that each chemical element content forms a complete feature combination with the corresponding comprehensive feature value. All such feature combinations of the samples are integrated to form a set containing double information of chemical composition and appearance features to obtain a joint feature set of the zinc-aluminum-magnesium plated layer sample.

[0114] When screening the subset of features in the joint feature set that are highly correlated with the corrosion resistance level of zinc-aluminum-magnesium coating samples, first determine the measured corrosion resistance level corresponding to each sample, and then analyze the correspondence between the numerical change of each feature in the joint feature set and the corrosion resistance level. When the value of a certain feature changes regularly, if the corrosion resistance level of the sample also shows a clear synchronous change trend, that is, when the value of the feature increases or decreases, the corrosion resistance level also increases or decreases accordingly, it indicates that the feature is highly correlated with the corrosion resistance level. All such features with clear synchronous change trends are selected, and the set formed is the feature subset of zinc-aluminum-magnesium coating samples.

[0115] When spatially compressing the feature subset to obtain the standard feature identifier of the zinc-aluminum-magnesium coating sample, the core connotation of each feature in the feature subset is first comprehensively analyzed to determine whether different features reflect the same essential law affecting corrosion resistance. Similar features reflecting the same law are merged, and duplicate or redundant information is removed, while key content that can reflect the core influencing law is retained. Then, according to the importance of each feature to corrosion resistance, the merged core features are arranged in an orderly manner to form a set of concise, concentrated features that can comprehensively cover the key characteristics of the sample. This feature combination is the standard feature identifier of the zinc-aluminum-magnesium coating sample.

[0116] The beneficial effects are that by normalizing the chemical composition data, deeply coupling the two types of feature sets, screening highly correlated features, and performing spatial compression operations, the representation format of chemical composition data is unified, and the deep integration of chemical composition and morphological features is achieved. At the same time, redundant information is eliminated and core features are focused on. The resulting standard feature identifiers are accurate and concise, providing high-quality core node data for the subsequent construction of corrosion resistance knowledge graphs, and ensuring the accuracy and efficiency of subsequent similar sample matching and corrosion resistance prediction.

[0117] S4. Using the standard feature identifiers as nodes and the correlation between the measured corrosion resistance levels in the zinc-aluminum-magnesium coating samples as edges, construct a knowledge graph of the corrosion resistance of the zinc-aluminum-magnesium coating samples.

[0118] In this embodiment of the invention, the step of constructing a corrosion resistance knowledge graph of the zinc-aluminum-magnesium coating samples, using the standard feature identifiers as nodes and the correlation relationships of the measured corrosion resistance levels in the zinc-aluminum-magnesium coating samples as edges, includes:

[0119] Spatial clustering analysis was performed on the standard feature identifiers to obtain feature similarity groups of the zinc-aluminum-magnesium coating samples;

[0120] The standard feature identifiers are mapped to nodes, and the nodes are grouped according to feature similarity to establish the topological connection relationship of the nodes, thereby obtaining the initial spectral structure of the zinc-aluminum-magnesium coating sample.

[0121] According to the correlation strength of the measured corrosion resistance grade of the zinc-aluminum-magnesium plated layer sample, the edge of the initial graph structure is weighted to obtain an intermediate graph structure of the zinc-aluminum-magnesium plated layer sample.

[0122] The intermediate graph structure is subjected to graph structure optimization to eliminate isolated nodes in the intermediate graph structure, thereby obtaining the corrosion resistance knowledge graph of the zinc-aluminum-magnesium plated layer sample.

[0123] When performing spatial clustering analysis on the standard feature identifiers of the zinc-aluminum-magnesium plated layer sample, the core feature content contained in each standard feature identifier is extracted one by one, and the core features of any two standard feature identifiers are comprehensively compared to count the coincidence of the feature types and feature value trends of the two, and when the core features of the two standard feature identifiers coincide to reach a set unified determination standard, the two standard feature identifiers are classified into the same group, and the two-by-two comparison and classification of all standard feature identifiers are completed in the same comparison manner, and finally a plurality of sets of similar feature attributes are formed to obtain the feature similarity grouping of the zinc-aluminum-magnesium plated layer sample.

[0124] Each standard feature identifier of the zinc-aluminum-magnesium plated layer sample corresponds to an independent node, and each node uniquely represents a standard feature identifier, and then according to the feature similarity grouping result, all nodes in the same group are connected with each other by straight lines, and the nodes of different groups are not connected for the time being, and the initial association relationship between the nodes is determined in this way to form a basic structure containing nodes and initial connection lines between the nodes, thereby obtaining the initial graph structure of the zinc-aluminum-magnesium plated layer sample.

[0125] The measured corrosion resistance grade corresponding to each zinc-aluminum-magnesium plated layer sample is collected, the measured corrosion resistance grades of the samples corresponding to the two nodes connected by each connection line in the initial graph structure are analyzed, if the measured corrosion resistance grades of the two nodes are completely consistent, the correlation strength of the two is the highest, if the grades are similar, the correlation strength is medium, and if the grades are quite different, the correlation strength is the lowest, according to different correlation strength grades, each connection line is marked with a corresponding weight identifier, and the weight identifier directly reflects the correlation closeness of the corrosion resistance grades between the nodes, after the weight marking of all connection lines is completed, the intermediate graph structure of the zinc-aluminum-magnesium plated layer sample is obtained.

[0126] When optimizing the intermediate graph structure of zinc-aluminum-magnesium coating samples, all nodes in the graph are thoroughly examined to identify isolated nodes that are not connected to any other nodes. For each isolated node, its corresponding standard feature identifier is re-compared with the core features of all feature similarity groups to find the group with the highest degree of overlap of core features. A weakly weighted identifier is then established between the isolated node and any node in the group to ensure that all nodes in the graph have at least one connection. After eliminating all isolated nodes, a knowledge graph of the corrosion resistance of zinc-aluminum-magnesium coating samples with a complete structure and reasonable node associations is obtained.

[0127] The beneficial effects are that spatial clustering analysis clarifies feature similarity groups, establishes node topological connections based on groups to form an initial graph, assigns edge weights based on measured corrosion resistance levels, and further optimizes and eliminates isolated nodes. The final corrosion resistance knowledge graph has clear node association logic, weight assignments that conform to actual corrosion resistance association rules, and a complete and non-redundant structure. It provides a stable and reliable graph foundation for subsequent matching of similar samples of the coating to be predicted and the derivation of corrosion resistance levels, effectively ensuring the accuracy and efficiency of the subsequent prediction process.

[0128] S5. Input the original analysis data of the zinc-aluminum-magnesium coating to be predicted into the corrosion resistance knowledge graph and identify similar sample nodes to the zinc-aluminum-magnesium coating to be predicted.

[0129] In this embodiment of the invention, the step of inputting the original analytical data of the zinc-aluminum-magnesium coating to be predicted into the corrosion resistance knowledge graph and identifying similar sample nodes to the zinc-aluminum-magnesium coating to be predicted includes:

[0130] The original analytical data of the zinc-aluminum-magnesium coating to be predicted is mapped to a feature space consistent with the zinc-aluminum-magnesium coating sample to obtain a temporary feature identifier of the zinc-aluminum-magnesium coating to be predicted.

[0131] The temporary feature identifier is compared with the standard feature identifier in the corrosion resistance knowledge graph in multiple dimensions to obtain candidate similar nodes of the zinc-aluminum-magnesium coating to be predicted.

[0132] Based on the topology of the corrosion resistance knowledge graph, the candidate similar nodes are expanded to obtain the associated nodes of the zinc-aluminum-magnesium coating to be predicted.

[0133] The confidence level of the associated nodes is evaluated, and nodes with high confidence levels are selected as similar sample nodes for the zinc-aluminum-magnesium coating to be predicted.

[0134] When mapping the original analysis data of the zinc-aluminum-magnesium coating to be predicted into the feature space consistent with the zinc-aluminum-magnesium coating sample, the original analysis data of the coating to be predicted, including chemical composition data and microstructure-related data, is extracted first, and the original chemical composition data is normalized according to the standard when the zinc-aluminum-magnesium coating sample is processed, and the original microstructure data is subjected to image segmentation, feature coupling and other operations consistent with the sample in sequence to obtain the comprehensive morphology features of the coating to be predicted. Then, the normalized chemical composition data and the comprehensive morphology features are integrated according to the feature type, classification rule and arrangement order of the sample feature space, so that the feature expression format and dimensionality of the coating to be predicted are completely matched with the feature space of the sample. Finally, the formed feature combination is the temporary feature identifier of the zinc-aluminum-magnesium coating to be predicted.

[0135] When comparing the temporary feature identifier with the standard feature identifier in the corrosion resistance knowledge graph in multiple dimensions, the comparison is carried out from three dimensions of feature type, feature value trend and core feature connotation, and the corresponding features of the temporary feature identifier and each standard feature identifier are compared one by one. The numerical coincidence, feature change trend consistency and core feature reflected rule coincidence of the two in the same feature type are counted. When the comprehensive coincidence degree of the three reaches the set unified determination standard, the graph node corresponding to the standard feature identifier is the candidate similar node of the zinc-aluminum-magnesium coating to be predicted. After the comparison of all standard feature identifiers is completed, the candidate similar node set is obtained by summarizing.

[0136] When the candidate similar nodes are expanded in the field based on the topological structure of the corrosion resistance knowledge graph, all topological connection relationships established in the graph for each candidate similar node are first determined, and all nodes directly connected to each candidate similar node through a connection line are found. These nodes have an explicit topological association with the candidate similar node, indicating that the standard feature identifier corresponding to the node is associated with the feature of the candidate similar node. All directly connected nodes of the candidate similar nodes are included in the associated node set together with the candidate similar nodes to obtain the associated nodes of the zinc-aluminum-magnesium coating to be predicted.

[0137] When evaluating the confidence of the associated nodes, two core dimensions are combined for determination. One is the multi-dimensional similarity coincidence degree of the standard feature identifier corresponding to the associated node and the temporary feature identifier, and the other is the weight identifier size of the connection line between the associated node and the candidate similar node. The higher the similarity coincidence degree and the greater the connection line weight identifier, the stronger the reliability of the associated node and the higher the confidence. According to the unified confidence determination standard, the nodes that meet the requirements in terms of coincidence degree and weight identifier are selected as the similar sample nodes of the zinc-aluminum-magnesium coating to be predicted.

[0138] The beneficial effect is that by mapping the to-be-predicted coating data to a unified feature space, the comparability with sample features is ensured, the multi-dimensional similarity comparison accurately locks the initial candidate node, the domain expansion based on the topological structure expands the correlation range, the confidence evaluation effectively filters out reliable nodes, and finally the similar sample nodes have both similarity and correlation, providing a precise and reliable basis for the subsequent derivation of corrosion resistance grade and ensuring the scientificity and accuracy of the subsequent prediction process.

[0139] S6, connecting the similar sample nodes into a relationship path of the to-be-predicted zinc-aluminum-magnesium coating, and performing traversal exploration along the relationship path to obtain a corrosion resistance prediction grade corresponding to the to-be-predicted zinc-aluminum-magnesium coating.

[0140] In the embodiment of the present application, the connecting the similar sample nodes into a relationship path of the to-be-predicted zinc-aluminum-magnesium coating comprises:

[0141] According to the spatial distribution of the similar sample nodes in the corrosion resistance knowledge graph, an initial path framework of the to-be-predicted zinc-aluminum-magnesium coating is established;

[0142] Based on the similarity of the similar sample nodes and the correlation degree of the corrosion resistance grade in the corrosion resistance knowledge graph, the paths in the initial path framework are prioritized to obtain a priority path framework of the to-be-predicted zinc-aluminum-magnesium coating;

[0143] A multi-path search strategy is adopted to perform path exploration on the priority path framework to obtain a candidate path of the to-be-predicted zinc-aluminum-magnesium coating;

[0144] The coherence of the candidate path is verified to obtain a relationship path of the to-be-predicted zinc-aluminum-magnesium coating.

[0145] The traversal exploration along the relationship path to obtain a corrosion resistance prediction grade corresponding to the to-be-predicted zinc-aluminum-magnesium coating comprises:

[0146] Starting from the starting node of the relationship path, the similar sample nodes are prioritized according to the path weight of the relationship path to obtain the corrosion resistance grade characteristics of the similar sample nodes;

[0147] The corrosion resistance grade characteristics are integrated, and the feature conflicts between the similar sample nodes are eliminated to obtain a corrosion resistance grade characteristic set of the to-be-predicted zinc-aluminum-magnesium coating;

[0148] The corrosion resistance grade characteristic set is trend analyzed to obtain a corrosion resistance prediction grade of the to-be-predicted zinc-aluminum-magnesium coating.

[0149] When establishing the initial path framework of the zinc-aluminum-magnesium coating to be predicted according to the spatial distribution of similar sample nodes in the corrosion resistance knowledge graph, the specific position of each similar sample node in the graph and the existing topological connection relationship between the nodes are first determined, all similar sample nodes are marked and positioned in the graph, and then the potential path directions that can be extended from different sample nodes are preliminarily sorted out according to the distribution density and connection tightness of the nodes. The potential path directions are combined with the actual connection between the nodes to construct a basic path set containing all possible connection modes, thereby forming the initial path framework of the zinc-aluminum-magnesium coating to be predicted.

[0150] When performing priority evaluation on the paths in the initial path framework based on the similarity of similar sample nodes and the correlation degree of the corrosion resistance level in the corrosion resistance knowledge graph, the similarity of all similar sample nodes on each path to the temporary characteristic identifier of the coating to be predicted is first determined, and then the correlation tightness of the corrosion resistance level corresponding to the nodes on each path to the core influencing factors of the coating corrosion resistance is confirmed. The similarity sum and the correlation sum of each path are comprehensively considered, the path with higher similarity sum and tighter correlation sum has higher priority, and all paths in the initial path framework are sorted according to this unified standard. The paths with high priority are selected to form a new framework, thereby obtaining the priority path framework of the zinc-aluminum-magnesium coating to be predicted.

[0151] When performing path exploration on the priority path framework by using the multi-path search strategy, the exploration starts from the starting point of each high-priority path in the priority path framework and extends in different directions along the connection between the nodes. The connection order and corresponding path weight of each node are recorded during the extension process to ensure that the exploration process covers all possible extension directions of the high-priority paths in the framework without missing any complete path with effective connection. All complete paths that meet the high-priority standard and are explored are collected and summarized to obtain the candidate paths of the zinc-aluminum-magnesium coating to be predicted.

[0152] When verifying the coherence of the candidate paths to obtain the relationship path of the zinc-aluminum-magnesium coating to be predicted, the connection logic of the nodes on each candidate path is checked one by one to confirm whether the connection between adjacent nodes on the path has a reasonable weight identifier, whether the characteristic attributes corresponding to the nodes have a continuous correlation trend, and whether there is any unrelated node jump or broken connection. At the same time, it is ensured that the corrosion resistance level correlation direction of the path from the starting point to the ending point is consistent and there is no logical contradiction. After verification, all coherent and logically defect-free candidate paths are retained to obtain the relationship path of the zinc-aluminum-magnesium coating to be predicted.

[0153] When the corrosion resistance grade characteristics of the similar sample nodes are obtained by preferentially traversing the similar sample nodes according to the path weight of the relationship path from the starting node of the relationship path, the starting node of each relationship path is determined as a traversal starting point, the traversal order is determined according to the weight identification of the connection between the nodes on the path, the next node corresponding to the connection with a larger weight identification is preferentially traversed, and the measured corrosion resistance grade corresponding to each traversed similar sample node is recorded one by one in the traversal process. The recorded measured corrosion resistance grade is the corrosion resistance grade characteristic of the similar sample node.

[0154] When the corrosion resistance grade characteristic set of the zinc-aluminum-magnesium coating to be predicted is obtained by integrating the corrosion resistance grade characteristics and eliminating the characteristic conflicts between the similar sample nodes, the corrosion resistance grade characteristics obtained by traversal are first classified and summarized according to the characteristic types, then it is checked whether there is a conflict of inconsistent grades for the same type of characteristics corresponding to different similar sample nodes, when a conflict occurs, the grade characteristic corresponding to the node with high similarity and large path weight is retained, the low-priority conflicting characteristic is eliminated, and finally a conflict-free grade characteristic set is formed, and the corrosion resistance grade characteristic set of the zinc-aluminum-magnesium coating to be predicted is obtained.

[0155] When the corrosion resistance prediction grade of the zinc-aluminum-magnesium coating to be predicted is obtained by trend analysis on the corrosion resistance grade characteristic set, the frequency of each type of corrosion resistance grade in the characteristic set is counted, the concentration trend of the grade distribution is analyzed, and the corrosion resistance grade corresponding to the highest proportion and the high-similarity and high-weight node is determined by combining the correlation between the grade in the characteristic set and the path weight and the node similarity. The grade is the most representative core grade in the characteristic set, and the core grade is determined as the corrosion resistance prediction grade of the zinc-aluminum-magnesium coating to be predicted.

[0156] The beneficial effects are that the corrosion resistance prediction grade is accurately locked by gradually constructing a path framework, screening high-priority paths, verifying path continuity, traversing nodes according to weights, fully integrating and eliminating conflicting corrosion resistance grade characteristics, the whole process is logically coherent and progressive, the effective information of the similar sample nodes is fully utilized, the interference of invalid data and conflicting characteristics is avoided, the accuracy and reliability of the corrosion resistance prediction grade are ensured, and a precise basis is provided for performance evaluation of the zinc-aluminum-magnesium coating to be predicted.

[0157] As shown in Figure 2 , it is a functional module diagram of a zinc-aluminum-magnesium coating corrosion resistance prediction system provided by an embodiment of the present application.

[0158] The zinc-aluminum-magnesium coating corrosion resistance prediction system 100 can be installed in an electronic device. According to the functions implemented, the zinc-aluminum-magnesium coating corrosion resistance prediction system 100 can include a data partition module 101, a feature coupling module 102, a standard feature identification generation module 103, a corrosion resistance graph construction module 104, a similar node identification module 105, and a corrosion resistance grade prediction module 106. The modules of the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, which are stored in the memory of the electronic device.

[0159] In the present embodiment, the functions of each module / unit are as follows:

[0160] The data partition module 101 is configured to segment the microstructure image of the zinc-aluminum-magnesium coating sample into different analysis levels and identify different material region data on the analysis levels to obtain structure partition data of the zinc-aluminum-magnesium coating sample.

[0161] The feature coupling module 102 is configured to perform multi-dimensional coupling on the geometric shape features and spatial distribution features in the structure partition data to obtain a comprehensive topographic feature set of the zinc-aluminum-magnesium coating sample.

[0162] The standard feature identification generation module 103 is configured to fuse the chemical composition data of the zinc-aluminum-magnesium coating sample and the comprehensive topographic feature set and perform feature dimension reduction processing on the fused feature set to obtain a standard feature identification of the zinc-aluminum-magnesium coating sample.

[0163] The corrosion resistance graph construction module 104 is configured to construct a corrosion resistance knowledge graph of the zinc-aluminum-magnesium coating sample by taking the standard feature identification as a node and the correlation between the measured corrosion resistance grades in the zinc-aluminum-magnesium coating sample as an edge.

[0164] The similar node identification module 105 is configured to input the original analysis data of the zinc-aluminum-magnesium coating to be predicted into the corrosion resistance knowledge graph and identify similar sample nodes of the zinc-aluminum-magnesium coating to be predicted.

[0165] The corrosion resistance grade prediction module 106 is configured to connect the similar sample nodes into a relationship path of the zinc-aluminum-magnesium coating to be predicted and perform traversal exploration along the relationship path to obtain a corrosion resistance prediction grade corresponding to the zinc-aluminum-magnesium coating to be predicted.

[0166] In the several embodiments provided in the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and for example, the division of the modules is only a logical functional division, and other division methods can be used in actual implementation.

[0167] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, i.e., may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0168] In addition, each functional module in each embodiment of the application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware, or in the form of hardware plus software functional module.

[0169] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0170] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology and application system for using digital computers or computer-controlled machines to simulate, extend and expand human intelligence, perceive environment, acquire knowledge and use knowledge to obtain optimal results.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for predicting corrosion resistance of a zinc-aluminum-magnesium plated layer, characterized by, The method comprises: S1, dividing the microstructure image of the zinc-aluminum-magnesium plated layer sample into different analysis levels, and identifying different material region data on the analysis levels to obtain structure partition data of the zinc-aluminum-magnesium plated layer sample; S2, coupling the geometric morphological features and spatial distribution features in the structure partition data in multiple dimensions to obtain a comprehensive topographic feature set of the zinc-aluminum-magnesium plated layer sample; S3, fusing the chemical composition data of the zinc-aluminum-magnesium plated layer sample and the comprehensive topographic feature set, and performing feature dimension reduction processing on the fused feature set to obtain a standard feature identifier of the zinc-aluminum-magnesium plated layer sample; S4, taking the standard feature identifier as a node, and taking the correlation relationship of the measured corrosion resistance grade in the zinc-aluminum-magnesium plated layer sample as an edge to construct a corrosion resistance knowledge graph of the zinc-aluminum-magnesium plated layer sample; S5, inputting the original analysis data of the zinc-aluminum-magnesium plated layer to be predicted into the corrosion resistance knowledge graph, and identifying similar sample nodes of the zinc-aluminum-magnesium plated layer to be predicted; S6, connecting the similar sample nodes into a relationship path of the zinc-aluminum-magnesium plated layer to be predicted, and traversing and exploring along the relationship path to obtain a corresponding corrosion resistance prediction grade of the zinc-aluminum-magnesium plated layer to be predicted.

2. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The method comprises: Collecting the microstructure image of the zinc-aluminum-magnesium plated layer sample, and performing multi-scale Gaussian filtering processing on the microstructure image to obtain a multi-scale image set of the zinc-aluminum-magnesium plated layer sample; Performing boundary feature recognition on the filtering region between the phases in the multi-scale image set to obtain phase boundary distribution information of the zinc-aluminum-magnesium plated layer sample; According to the phase boundary distribution information, the multi-scale image set is segmented by using a region growing technique to obtain independent partitions of the zinc-aluminum-magnesium plated layer sample; Performing morphological optimization processing on the independent partitions to obtain material regions of the zinc-aluminum-magnesium plated layer sample; Quantifying the geometric features and spatial relationship features in the material regions to obtain structure partition data of the zinc-aluminum-magnesium plated layer sample.

3. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The method comprises: Analyzing the region contour information in the structure partition data to obtain geometric descriptors of the zinc-aluminum-magnesium plated layer sample; According to the region spatial coordinates and spatial topological structure in the structure partition data, a region relationship network of the zinc-aluminum-magnesium plated layer sample is constructed; Identifying the region distribution mode in the region relationship network to obtain a spatial distribution feature descriptor of the zinc-aluminum-magnesium plated layer sample; Standardizing and integrating the geometric descriptors and the spatial distribution feature descriptors to obtain a geometric morphological feature set and a spatial distribution feature set of the zinc-aluminum-magnesium plated layer sample; Performing tensor synthesis on the geometric morphological feature set and the spatial distribution feature set to obtain a geometric feature vector and a spatial feature vector of the zinc-aluminum-magnesium plated layer sample; The geometric feature vector and the spatial feature vector are feature-coupled to obtain a comprehensive feature value of the zinc-aluminum-magnesium coating sample, and the comprehensive feature value is integrated to obtain a comprehensive morphology feature set of the zinc-aluminum-magnesium coating sample.

4. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 3, characterized in that, The formula for calculating the comprehensive feature value is: ; wherein, denotes the comprehensive feature value, denotes the mean value of the geometric feature vectors, denotes the geometric feature vector, denotes the mean value of the geometric feature vectors, denotes the spatial feature vector, denotes the spatial feature vector, denotes the mean value of the spatial feature vectors, denotes the preset weight coefficient, denotes the number of all feature vectors.

5. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The chemical composition data of the zinc-aluminum-magnesium coating sample and the comprehensive morphology feature set are fused, and feature dimension reduction processing is performed on the fused feature set to obtain a standard feature identifier of the zinc-aluminum-magnesium coating sample, including: The chemical composition data of the zinc-aluminum-magnesium coating sample is normalized to obtain a standard composition set of the zinc-aluminum-magnesium coating sample; The standard composition set and the comprehensive morphology feature set are deeply coupled to obtain a joint feature set of the zinc-aluminum-magnesium coating sample; A feature subset with a high correlation degree to the corrosion resistance grade in the zinc-aluminum-magnesium coating sample is screened from the joint feature set; The feature subset is spatially compressed to obtain a standard feature identifier of the zinc-aluminum-magnesium coating sample.

6. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The standard feature identifier is taken as a node, and a correlation between the measured corrosion resistance grade in the zinc-aluminum-magnesium coating sample is taken as an edge to construct a corrosion resistance knowledge graph of the zinc-aluminum-magnesium coating sample, including: Spatial clustering analysis is performed on the standard feature identifier to obtain a feature similarity grouping of the zinc-aluminum-magnesium coating sample; The standard feature identifier is mapped to a node, and a topological connection relationship of the node is established according to the feature similarity grouping to obtain an initial graph structure of the zinc-aluminum-magnesium coating sample; According to the correlation strength of the measured corrosion resistance grade in the zinc-aluminum-magnesium coating sample, a weight amplitude is given to the edge of the initial graph structure to obtain an intermediate graph structure of the zinc-aluminum-magnesium coating sample; The intermediate graph structure is graph structure optimized to eliminate isolated nodes in the intermediate graph structure to obtain the corrosion resistance knowledge graph of the zinc-aluminum-magnesium coating sample.

7. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The original analysis data of the zinc-aluminum-magnesium coating to be predicted is input into the corrosion resistance knowledge graph, and a similar sample node of the zinc-aluminum-magnesium coating to be predicted is identified, including: The original analysis data of the zinc-aluminum-magnesium coating to be predicted is mapped to a feature space consistent with the zinc-aluminum-magnesium coating sample to obtain a temporary feature identifier of the zinc-aluminum-magnesium coating to be predicted; The temporary feature identifier is compared with the standard feature identifier in the corrosion resistance knowledge graph in multiple dimensions to obtain a candidate similar node of the zinc-aluminum-magnesium coating to be predicted; Based on the topological structure of the corrosion resistance knowledge graph, the candidate similar node is expanded in the field to obtain an associated node of the zinc-aluminum-magnesium coating to be predicted; The confidence of the associated node is evaluated, and the node with high confidence is taken as the similar sample node of the zinc-aluminum-magnesium coating to be predicted.

8. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The similar sample nodes are connected to form a relationship path of the zinc-aluminum-magnesium coating to be predicted, including: An initial path framework of the zinc-aluminum-magnesium coating to be predicted is established according to the spatial distribution of the similar sample nodes in the corrosion resistance knowledge graph; Based on the correlation between the similarity of the similar sample node and the corrosion resistance level in the corrosion resistance knowledge graph, the path in the initial path framework is evaluated in priority to obtain a priority path framework of the zinc-aluminum-magnesium coating to be predicted; A multi-path search strategy is adopted to explore the priority path framework to obtain a candidate path of the zinc-aluminum-magnesium coating to be predicted; The coherence of the candidate path is verified to obtain a relationship path of the zinc-aluminum-magnesium coating to be predicted.

9. A method of predicting the corrosion resistance of a zinc-aluminum-magnesium coating as defined in claim 1, characterized in that, The traversal exploration along the relationship path obtains a corrosion resistance prediction level corresponding to the zinc-aluminum-magnesium coating to be predicted, including: Starting from the starting node of the relationship path, the similar sample nodes are traversed in priority according to the path weight of the relationship path to obtain the corrosion resistance level characteristics of the similar sample nodes; The corrosion resistance level characteristics are integrated, and the feature conflicts between the similar sample nodes are eliminated to obtain a corrosion resistance level characteristic set of the zinc-aluminum-magnesium coating to be predicted; The corrosion resistance level characteristic set is analyzed in trend to obtain the corrosion resistance prediction level of the zinc-aluminum-magnesium coating to be predicted. 10.A zinc-aluminum-magnesium coating corrosion resistance prediction system for implementing the zinc-aluminum-magnesium coating corrosion resistance prediction method of claim 1, the system comprising: a data partition module configured to segment a microstructure image of a zinc-aluminum-magnesium coating sample into different analysis levels, and identify different material region data on the analysis levels to obtain structure partition data of the zinc-aluminum-magnesium coating sample; a feature coupling module configured to perform multi-dimensional coupling on geometric morphology features and spatial distribution features in the structure partition data to obtain a comprehensive topography feature set of the zinc-aluminum-magnesium coating sample; a standard feature identification generation module configured to fuse chemical composition data of the zinc-aluminum-magnesium coating sample and the comprehensive topography feature set, and perform feature dimension reduction processing on the fused feature set to obtain a standard feature identification of the zinc-aluminum-magnesium coating sample; a corrosion resistance graph construction module configured to construct a corrosion resistance knowledge graph of the zinc-aluminum-magnesium coating sample by taking the standard feature identification as a node and taking the correlation relationship of the measured corrosion resistance level in the zinc-aluminum-magnesium coating sample as an edge; a similar node identification module configured to input original analysis data of a zinc-aluminum-magnesium coating to be predicted into the corrosion resistance knowledge graph and identify similar sample nodes of the zinc-aluminum-magnesium coating to be predicted; a corrosion resistance level prediction module configured to connect the similar sample nodes into a relationship path of the zinc-aluminum-magnesium coating to be predicted, and perform traversal exploration along the relationship path to obtain a corrosion resistance prediction level corresponding to the zinc-aluminum-magnesium coating to be predicted.

Images