A pipeline corrosion product multi-dimensional feature extraction and corrosion state evaluation method and system

Abstract

The application discloses a kind of pipeline corrosion product multidimensional feature extraction and corrosion state evaluation method and system for water supply pipeline corrosion evaluation technical field, including obtaining the pipeline corrosion product in several pipelines;The micro-characterization data of the pipeline corrosion product is obtained, the micro-characterization data is standardized to obtain micro standard data, and the micro standard data is screened to obtain micro verification data;The common factor data and corresponding factor loading matrix that can represent the pipeline corrosion state are extracted from the micro verification data that meet the requirements;Based on the preset factor direction, the common factor is divided into positive factor and negative factor;According to the factor loading matrix, the positive factor and the negative factor direction rule are calculated to obtain comprehensive evaluation data, and the pipeline corrosion degree is judged by combining the comprehensive evaluation data and the preset value, so as to realize the comprehensive detection and accurate quantitative evaluation of pipeline corrosion state.

Description

Technical Field

[0001] This invention relates to the field of corrosion evaluation technology for water supply pipelines, specifically a method and system for extracting multi-dimensional features of pipeline corrosion products and evaluating corrosion status. Background Technology

[0002] Corrosion of the inner walls of water supply pipelines is a common problem in water supply systems, often leading to pipeline leaks, water quality deterioration, and reduced water supply efficiency. Existing pipeline corrosion evaluation methods, such as a stability index system and evaluation method for water supply networks (authorization announcement number CN108764594B), provide an index system that combines water quality and pipeline surface characteristics, and uses segmented thresholds to judge the health status of pipelines. However, the evaluation system is simplistic and the classification principle is coarse, failing to accurately define the pipeline condition. A detection method and system for internal corrosion in long-distance pipelines (publication number CN119400271A) constructs a multi-level corrosion evaluation system, combining factor weights and membership matrices to determine the corrosion degree level, but the weight allocation has subjective biases and limited applicability.

[0003] However, the current methods for determining the corrosion status of water supply pipelines are not clear. Existing detection and evaluation methods often cannot fully and accurately reflect the corrosion status of pipelines. Traditional detection methods usually rely on single physical or chemical indicators, such as pipeline wall thickness measurement, potential and current monitoring, and water quality parameter analysis, which are difficult to fully reveal the complexity and multi-dimensional characteristics of pipeline corrosion. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for extracting multi-dimensional features of pipeline corrosion products and evaluating corrosion status, in order to solve the problem that the method for determining the corrosion status of water supply pipelines is not yet clear, and existing detection and evaluation methods often cannot fully and accurately reflect the corrosion status of pipelines.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides a method for multi-dimensional feature extraction of pipeline corrosion products and evaluation of corrosion status, characterized in that it includes:

[0007] Obtain pipeline corrosion products from several pipelines;

[0008] Microscopic characterization data of the pipeline corrosion products are obtained, and the microscopic characterization data are standardized to obtain microscopic standard data. Based on the suitability test of factor analysis, the microscopic standard data are screened to obtain microscopic verification data.

[0009] The micro-validation data is subjected to dimensionality reduction analysis, and common factors of several targets are extracted and a factor loading matrix is ​​constructed. The factor loading matrix includes the loadings on the common factors of several targets. The common factors are divided into positive factors and negative factors based on a preset factor direction.

[0010] The factor loading matrix is ​​coupled with the positive and negative factors to calculate the comprehensive evaluation data of the pipeline corrosion status. If the comprehensive evaluation data is less than the preset value, the pipeline is output as having a corrosion risk; otherwise, the pipeline is output as being in a healthy state.

[0011] As a further aspect of the present invention: obtaining the microscopic characterization data of the pipeline corrosion products includes;

[0012] The microscopic characterization data includes elemental data, pore volume data, average pore diameter data, specific surface area data, and crystal structure data of pipeline corrosion products.

[0013] The elemental index data includes the relative content of iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese and chlorine in the pipeline corrosion products;

[0014] The method for obtaining the elemental index data includes at least one of X-ray fluorescence spectroscopy, X-ray photoelectron spectroscopy, inductively coupled plasma spectroscopy, and scanning electron microscopy-energy dispersive spectroscopy.

[0015] The methods for acquiring the pore volume data, average pore diameter data, specific surface area data, and crystal structure data include at least one of fully automated specific surface area and porosity analysis, X-ray diffraction, and scanning electron microscopy.

[0016] As a further aspect of the present invention: the elemental index data, pore volume data, average pore diameter data, specific surface area data and crystal structure data are standardized to obtain microscopic standard data;

[0017] The standardization process includes data transformation and data standardization;

[0018] The data transformation includes performing a central logarithmic ratio transformation on the relative content data to eliminate the pseudo-correlation between iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese and chlorine in pipeline corrosion products caused by the constant total amount.

[0019] The data standardization includes Z-score standardization of the micro-standard data, and the Z-score standardization formula is:

[0020]

[0021] Where x is the original data representing the microscopic characteristics, x' is the standardized data, mean(x) is the sample mean, and σ is the sample standard deviation.

[0022] As a further aspect of the present invention: the suitability test based on factor analysis filters the micro-standard data to obtain micro-validation data, including;

[0023] The data correlation matrix is ​​obtained from the micro-standard data, and the micro-validation data is obtained by screening the micro-standard data based on the suitability test of factor analysis.

[0024] A data matrix is ​​constructed by standardizing the micro-standard data, and the data matrix is ​​transformed into a standardized data matrix, which is X. A data correlation matrix is ​​obtained based on the standardized data matrix.

[0025] The data correlation matrix is ​​as follows:

[0026]

[0027] Where X T Let X be the transpose of the standardized data matrix X, and n be the number of samples.

[0028] The suitability tests for the factor analysis include the KMO test and the Bartlett test for sphericity, and the micro-standard data are screened to obtain micro-validation data based on the preset values ​​of the KMO test and the Bartlett test for sphericity.

[0029] As a further aspect of the present invention: the microscopic verification data is subjected to dimensionality reduction analysis and several common factors of the targets are extracted and a factor loading matrix is ​​constructed. The factor loading matrix includes loadings on the common factors of several targets, including:

[0030] The dimensionality reduction analysis method includes principal component analysis, which includes:

[0031] Perform eigenvalue decomposition on the data correlation matrix R:

[0032] R = P∧P T ;

[0033] Where P is: the eigenvector matrix, P T ∧ is the transpose of the eigenvector matrix, and ∧ is the diagonal eigenvalue matrix, where the eigenvalues ​​in ∧ are the eigenvalues.

[0034] Based on the criterion that the eigenvalues ​​are greater than 1, the common factors of the first few objectives are selected, and the number of common factors is K. A factor loading matrix L of the common factors of the first few objectives is then constructed.

[0035]

[0036] The proportion of explained variance of the i-th common factor among the K common factors:

[0037]

[0038] Where λ i Let be the eigenvalue corresponding to the i-th common factor, and k be the number of common factors extracted.

[0039] As a further aspect of the present invention: the common factors of the plurality of objectives include a first common factor, a second common factor, and a third common factor;

[0040] Among them, the iron content is the dominant variable in the first common factor, and the loading signs of calcium and silicon are opposite to those of iron, so as to reflect the inverse relationship between metal corrosion and surface protective layer, and to characterize the degree of internal wall corrosion and the state of protective layer.

[0041] The second common factor uses the pore size parameter as the dominant variable to reflect the porosity characteristics of the corrosion product structure and to indicate the compactness and integrity of the corrosion layer.

[0042] The third common factor has manganese content as the dominant variable and is positively correlated with phosphorus content, reflecting the enrichment degree of exogenous elements in corrosion products and is used to evaluate the component migration characteristics during the corrosion process.

[0043] As a further aspect of the present invention: the common factors are divided into positive factors and negative factors based on a preset factor direction, including;

[0044] When the loading value of the iron element content in the first common factor is less than -0.5, and the loading values ​​of the calcium and silicon element contents in the first common factor are greater than 0.5, the common factor corresponding to the iron element is a positive factor.

[0045] When the loading value of the aperture data in the second common factor is less than -0.5, the common factor corresponding to the aperture data is a positive factor;

[0046] When the loading values ​​of the manganese and phosphorus elements in the third common factor are less than -0.5, the common factor corresponding to the manganese and phosphorus elements is a positive factor.

[0047] If the loading direction of the main contributing feature in the common factor is inconsistent with the preset factor direction, the corresponding common factor is divided into reverse factors according to the sign pair corresponding to the loading value in the common factor.

[0048] As a further aspect of the present invention: the comprehensive evaluation data of pipeline corrosion state calculated by coupling the factor loading matrix with positive and negative factors includes;

[0049] The factor score matrix F is calculated using regression analysis on the aforementioned common factor data.

[0050] F = R -1 L(L T R -1 L) -1 X

[0051] Where R is the data correlation matrix, L is the factor loading matrix, and X is the standardized data matrix.

[0052] Based on the coupling of the common factor terms F1, F2, F3 of the factor score matrix F with their corresponding common factor weights, the comprehensive evaluation score Y of the pipeline corrosion state is calculated:

[0053] Y = Y1 + Y2 + Y3

[0054] Wherein, the common factor weight is the proportion of the variance explained by the common factor, and Y1, Y2 and Y3 represent the weighted contribution values ​​of the first common factor, the second common factor and the third common factor, respectively;

[0055] If the overall evaluation score of the corrosion state is less than 0, the output pipeline is at risk of corrosion; otherwise, the output pipeline is in a healthy state.

[0056] As a further aspect of the present invention, it also includes: performing Spearman analysis or Pearson analysis on the comprehensive evaluation score of the corrosion state and existing pipeline health evaluation parameters to obtain correlation coefficient data. If the correlation coefficient data is higher than 0.9, it indicates that the dimensionality reduction analysis is effective. The existing pipeline health evaluation parameters are obtained by scoring the pipeline health based on four dimensions: actual images of the pipeline interior, the proportion of rust area, the thickness of corrosion nodules, the surface smoothness, and the integrity of the pipe wall.

[0057] Secondly, this invention provides a system for multi-dimensional feature extraction and corrosion status evaluation of pipeline corrosion products, the system comprising:

[0058] A sampling module, used to acquire pipeline corrosion products in several pipelines;

[0059] The screening module acquires the microscopic characterization data of the pipeline corrosion products, standardizes the microscopic characterization data to obtain microscopic standard data, and screens the microscopic standard data based on the suitability test of factor analysis to obtain microscopic verification data.

[0060] The analysis module performs dimensionality reduction analysis on the micro-validation data, extracts common factors of several targets, and constructs a factor loading matrix. The factor loading matrix includes the loadings on the common factors of several targets. The common factors are divided into positive factors and negative factors based on a preset factor direction.

[0061] The output module couples the factor loading matrix with the positive and negative factors and calculates the comprehensive evaluation data of the pipeline corrosion status. If the comprehensive evaluation data is less than the preset value, the output pipeline is considered to be at risk of corrosion; otherwise, the output pipeline is considered to be in a healthy state.

[0062] Compared with the prior art, the beneficial effects of the present invention are:

[0063] 1. In this invention, after acquiring and processing the microscopic characterization data of pipeline corrosion products, microscopic verification data is obtained by screening the microscopic standard data. Data that meets the requirements is selected first. Common factor data and its factor loading matrix that can represent the pipeline corrosion state are extracted from the processed microscopic verification data. Comprehensive evaluation data is calculated according to the direction rules of the factor loading matrix and positive and negative factors. Based on the comprehensive evaluation data, a comprehensive detection and accurate assessment of the pipeline corrosion state can be achieved. Through corrosion product feature extraction and data dimensionality reduction analysis, the pipeline corrosion state is quantitatively calculated, avoiding the subjective bias of traditional assessment methods, improving the scientificity and accuracy of corrosion monitoring, and achieving good results.

[0064] 2. In this invention, by using data dimensionality reduction and factor analysis, key corrosion features are quickly extracted and scores are obtained, thereby achieving efficient evaluation of pipeline corrosion status, optimizing pipeline maintenance strategies, reducing operational risks, and facilitating efficient analysis and optimized maintenance.

[0065] 3. This invention has strong applicability and can be applied to different pipe materials and environmental media. It can also be integrated with existing pipe network management systems to improve the intelligent level of pipe corrosion prevention and control and realize efficient operation and maintenance of water supply systems. Attached Figure Description

[0066] Figure 1 This is a schematic diagram of the method flow structure of the present invention;

[0067] Figure 2 This is a common factor extraction diagram for principal component analysis provided in Embodiment 1 of the present invention;

[0068] Figure 3 This is a correlation analysis chart between pipeline corrosion score and subjective health score based on actual pipeline images provided in Embodiment 1 of the present invention.

[0069] Figure 4 This is a correlation analysis diagram between pipeline corrosion score and subjective health score based on actual pipeline images provided in Comparative Example 1 of this invention. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example:

[0072] Please see Figure 1 In this embodiment of the invention, a method for multi-dimensional feature extraction of pipeline corrosion products and evaluation of corrosion status is characterized by comprising:

[0073] S1: Obtain pipeline corrosion products from several pipelines;

[0074] S2: Obtain microscopic characterization data of pipeline corrosion products, standardize the microscopic characterization data to obtain microscopic standard data, and screen the microscopic standard data based on the suitability test of factor analysis to obtain microscopic validation data.

[0075] S3: Perform dimensionality reduction analysis on the micro-validation data and extract common factors of several targets and construct a factor loading matrix. The factor loading matrix includes the loadings on the common factors of several targets. The common factors are divided into positive factors and negative factors based on the preset factor direction.

[0076] S4: Couple the factor loading matrix with the positive and negative factors to calculate the comprehensive evaluation data of the pipeline corrosion status. If the comprehensive evaluation data is less than the preset value, the output pipeline is at risk of corrosion; otherwise, the output pipeline is in a healthy state.

[0077] Specifically, this invention obtains microscopic standard data by acquiring and processing microscopic characterization data of pipeline corrosion products. Microscopic verification data is then obtained by screening the microscopic standard data and selecting the most suitable data. Common factor data and their factor loading matrices, representing the pipeline corrosion state, are extracted from the processed microscopic verification data. Comprehensive evaluation data is calculated based on the direction rules of the factor loading matrix and positive and negative factors. This comprehensive evaluation data enables comprehensive detection and accurate assessment of the pipeline corrosion state. Through corrosion product feature extraction and data dimensionality reduction analysis, the pipeline corrosion state is quantitatively calculated, avoiding the subjective bias of traditional assessment methods and improving the scientific rigor and accuracy of corrosion monitoring. The invention also demonstrates good performance.

[0078] Furthermore, by utilizing data dimensionality reduction and factor analysis, key corrosion features can be quickly extracted and scores obtained, enabling efficient evaluation of pipeline corrosion status, optimizing pipeline maintenance strategies, reducing operational risks, and facilitating efficient analysis and optimized maintenance.

[0079] Furthermore, this invention is highly applicable to different pipe materials and environmental media, and can be integrated with existing pipe network management systems to improve the intelligence level of pipe corrosion prevention and control, thereby achieving efficient operation and maintenance of water supply systems.

[0080] Preferably, microscopic characterization data of pipeline corrosion products are obtained, including:

[0081] Microscopic characterization data include elemental data, pore volume data, average pore diameter data, specific surface area (BET) data, and crystal structure data of pipeline corrosion products;

[0082] The elemental index data includes the relative content of iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese and chlorine in the pipeline corrosion products;

[0083] The methods for obtaining elemental index data include at least one of X-ray fluorescence spectroscopy, X-ray photoelectron spectroscopy, inductively coupled plasma spectroscopy, and scanning electron microscopy-energy dispersive spectroscopy.

[0084] The methods for acquiring pore volume data, average pore diameter data, specific surface area data, and crystal structure data include at least one of fully automated specific surface area and porosity analysis, X-ray diffraction, and scanning electron microscopy.

[0085] Specifically, the elemental index data, pore volume data, average pore diameter data, specific surface area data, and crystal structure data of pipeline corrosion products together constitute a comprehensive and detailed microscopic description of pipeline corrosion products. Standardized processing steps ensure that data obtained from different sources and measurement methods can be compared and analyzed on the same scale, improving the comparability of data and the accuracy of analysis results. Through the suitability test of factor analysis, the microscopic verification data that are most relevant to the pipeline corrosion state and have the highest information content are further screened out. Using data dimensionality reduction and factor analysis, key corrosion characteristics are quickly extracted and scores are obtained, achieving efficient evaluation of pipeline corrosion state, optimizing pipeline network maintenance strategies, and reducing operational risks.

[0086] Specifically, X-ray fluorescence spectroscopy can rapidly and accurately analyze the elemental composition of pipeline corrosion products, while X-ray photoelectron spectroscopy can provide more in-depth information on the chemical state of elements. X-ray diffraction is mainly used to analyze the crystal phase structure of corrosion products, while scanning electron microscopy-energy dispersive spectroscopy can directly observe the microscopic morphology of corrosion products and perform quantitative elemental analysis. Fully automated specific surface area and porosity analysis is a common method for determining pore volume, average pore diameter, and specific surface area. It utilizes the principle of gas adsorption to accurately measure these key parameters. X-ray diffraction and scanning electron microscopy also play important roles in acquiring crystal structure data, revealing the crystal structure and microscopic morphological characteristics of corrosion products; they can comprehensively and accurately obtain microscopic characterization data of pipeline corrosion products.

[0087] Preferably, the elemental index data, pore volume data, average pore size data, specific surface area data, and crystal structure data are standardized to obtain microscopic standard data;

[0088] Standardization processes include data transformation and data standardization;

[0089] Data transformation includes performing central logarithmic ratio transformation on the relative content data to eliminate spurious correlations among iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese and chlorine in pipeline corrosion products caused by constant total amount;

[0090] Data standardization includes Z-score standardization of micro-standard data. The Z-score standardization formula is:

[0091]

[0092] Where x is the original data representing the microscopic characteristics, x' is the standardized data, mean(x) is the sample mean, and σ is the sample standard deviation.

[0093] Specifically, the data correlation matrix is ​​used to analyze the correlation between various micro-level representation data, revealing their intrinsic connections. By standardizing these data, the dimensional differences between different data can be eliminated, making the data more comparable.

[0094] Z-score standardization is a commonly used data standardization method that transforms raw data into standard normal distribution data with a mean of 0 and a standard deviation of 1, making the data more concentrated and easier to analyze. Data transformation includes central logarithmic ratio transformation of relative content data to eliminate spurious correlations among iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese and chlorine in pipeline corrosion products caused by constant total amount. Data transformation also involves appropriate transformation of specific types of data to meet the requirements of subsequent analysis models.

[0095] Furthermore, the standardization process also includes at least one of min-max standardization, centralization, and scaling methods.

[0096] Preferably, the micro-validation data are obtained by screening the micro-standard data based on factor analysis suitability tests, including:

[0097] The data correlation matrix was obtained from the micro-standard data, and the micro-validation data was obtained by screening the micro-standard data based on the suitability test of factor analysis.

[0098] A data matrix is ​​constructed from the standardized micro-standard data. The data matrix is ​​then transformed into a standardized data matrix, which is X. The data correlation matrix is ​​obtained from the standardized data matrix.

[0099] The data correlation matrix is ​​as follows:

[0100]

[0101] Where X T Let X be the transpose of the standardized data matrix X, and n be the number of samples.

[0102] The suitability tests for factor analysis include the KMO test and the Bartlett test for sphericity. Based on the preset values ​​of the KMO test and the Bartlett test for sphericity, the micro-standard data are screened to obtain micro-validation data.

[0103] Specifically, suitability tests for factor analysis include the KMO test and Bartlett's test of sphericity;

[0104] KMO test:

[0105]

[0106] in For: the sum of squares of the pairwise Pearson correlation coefficients between the indicator variables, For: the sum of squares of the partial correlation coefficients;

[0107] Bartlett's test for sphericity:

[0108]

[0109] Where χ 2 Here, n is the sample size, and p is the number of indicator variables;

[0110] If KMO > 0.5 in the KMO test and P < 0.05 in the Bartlett's test of sphericity, then the requirements for factor analysis are met, and micro-validation data are obtained.

[0111] Dimensionality reduction analysis methods include principal component analysis (PCA), maximum likelihood method, and principal axis factor method; common factors extracted by principal component analysis include:

[0112] Perform eigenvalue decomposition on the data correlation matrix R:

[0113] R = P∧P T ;

[0114] Where P is: the eigenvector matrix, P T ∧ is the transpose of the eigenvector matrix, and ∧ is the diagonal eigenvalue matrix, where the eigenvalues ​​in ∧ are the eigenvalues.

[0115] The first k common factors are selected based on the criterion that the eigenvalues ​​are greater than 1, and the factor loading matrix L, representing the contribution of each original variable to the common factors, is calculated:

[0116]

[0117] The factor loading matrix can also be rotated using orthogonal rotation methods to enhance the clarity of attribution between dominant variables and common factors;

[0118] The percentage of variance explained by the i-th common factor:

[0119]

[0120] Where λ i Let be the eigenvalue corresponding to the i-th common factor, and k be the number of common factors extracted.

[0121] Preferably, the micro-validation data is subjected to dimensionality reduction analysis and several common factors of the targets are extracted and a factor loading matrix is ​​constructed. The factor loading matrix includes loadings on the common factors of the targets, including:

[0122] Dimensionality reduction analysis methods include principal component analysis, which includes:

[0123] Perform eigenvalue decomposition on the data correlation matrix R:

[0124] R = P∧P T ;

[0125] Where P is: the eigenvector matrix, P T ∧ is the transpose of the eigenvector matrix, and ∧ is the diagonal eigenvalue matrix, where the eigenvalues ​​in ∧ are the eigenvalues.

[0126] Based on the criterion that the eigenvalues ​​are greater than 1, select the common factors of the first few objectives, with the number of common factors being K, and construct the factor loading matrix L of the common factors of the first few objectives:

[0127]

[0128] The proportion of explained variance of the i-th common factor among the K common factors:

[0129]

[0130] Where λ i Let be the eigenvalue corresponding to the i-th common factor, and k be the number of common factors extracted.

[0131] Preferably, the common factors of the objectives include a first common factor, a second common factor, and a third common factor;

[0132] Among them, the iron content is the dominant variable in the first common factor, and the loading signs of calcium and silicon are opposite to those of iron, which is used to reflect the inverse relationship between metal corrosion and surface protective layer, and to characterize the degree of internal wall corrosion and the state of protective layer.

[0133] The second common factor uses the pore size parameter as the dominant variable to reflect the porosity characteristics of the corrosion product structure and to indicate the compactness and integrity of the corrosion layer.

[0134] The third common factor has manganese content as the dominant variable and is positively correlated with phosphorus content, reflecting the enrichment degree of exogenous elements in corrosion products and is used to assess the component migration characteristics during the corrosion process.

[0135] Preferably, the common factors are divided into positive factors and negative factors based on a preset factor direction, including:

[0136] When the loading value of iron content in the first common factor is less than -0.5, and the loading values ​​of calcium and silicon content in the first common factor are greater than 0.5, the common factor corresponding to iron is a positive factor.

[0137] When the loading value of the aperture data in the second common factor is less than -0.5, the common factor corresponding to the aperture data is a positive factor;

[0138] When the loading values ​​of manganese and phosphorus in the third common factor are less than -0.5, the common factors corresponding to manganese and phosphorus are positive factors.

[0139] If the loading direction of the main contributing feature in the common factor is inconsistent with the preset factor direction, the corresponding common factor is classified as a reverse factor according to the sign pair corresponding to the loading value in the common factor.

[0140] Specifically, if the loading direction of the main contributing feature in a common factor is opposite to the preset factor direction (i.e., the loading value is negative and the absolute value is greater than the preset threshold), then the common factor is considered a reverse factor. A reverse factor indicates that the trend of the feature represented by the common factor in corrosion products is contrary to expectations or conventional understanding, revealing the influence of special corrosion mechanisms or environmental factors on the corrosion process. By integrating information from both positive and reverse factors, a more comprehensive understanding of the multi-dimensional characteristics of pipeline corrosion products can be achieved, thereby accurately evaluating the corrosion state.

[0141] Preferably, the comprehensive evaluation data of pipeline corrosion status is calculated by coupling the factor loading matrix with positive and negative factors, including:

[0142] The factor score matrix F is calculated using regression analysis on the common factor data.

[0143] F = R -1 L(L T R -1 L) -1 X

[0144] Where R is the data correlation matrix, L is the factor loading matrix, and X is the standardized data matrix.

[0145] Based on the coupling of the common factor terms F1, F2, F3 of the factor score matrix F with their corresponding common factor weights, the comprehensive evaluation score Y of the pipeline corrosion state is calculated:

[0146] Y = Y1 + Y2 + Y3

[0147] Wherein, the common factor weight is the proportion of the variance explained by the common factor, and Y1, Y2 and Y3 represent the weighted contribution values ​​of the first common factor, the second common factor and the third common factor, respectively.

[0148] If the overall evaluation score of corrosion status is less than 0, the output pipeline is at risk of corrosion; otherwise, the output pipeline is in a healthy state.

[0149] For newly introduced sample data, the existing standardization method, factor loading matrix and direction rules can be directly substituted into the data to calculate the comprehensive evaluation score of corrosion status without reconstructing the factor structure.

[0150] Preferably, the method further includes: performing Spearman or Pearson analysis on the comprehensive evaluation score of corrosion status and existing pipeline health evaluation parameters to obtain correlation coefficient data. If the correlation coefficient data is higher than 0.9, it indicates that the dimensionality reduction analysis is effective. The existing pipeline health evaluation parameters are obtained by scoring the pipeline health from four dimensions: actual images of the pipeline interior, the proportion of rust area, the thickness of corrosion nodules, the surface smoothness, and the integrity of the pipe wall, which is the subjective health score of the pipeline.

[0151] Obtain pipeline corrosion products from several pipelines, including:

[0152] The pipeline corrosion products are obtained from the inner wall of the pipeline using a scraper, tweezers, or a sampling hammer. The sampling time for the pipeline corrosion products is limited to within 2 hours after the inner wall of the pipeline is separated from the water. After collection, the pipeline corrosion products are transported and stored at low temperature and dried at 60°C.

[0153] Specifically, during sampling, it is essential to ensure that tools such as scrapers, tweezers, or sampling hammers are clean and uncontaminated to avoid secondary contamination of the corrosion products. After sampling, the corrosion products should be labeled and recorded for subsequent analysis and research. Low-temperature transportation and preservation can slow down the chemical changes of the corrosion products and maintain their original state, while drying at 60°C can remove moisture, facilitating subsequent chemical analysis and testing.

[0154] Not shown in the figure, this invention provides a system for multi-dimensional feature extraction of pipeline corrosion products and corrosion state evaluation, the system comprising:

[0155] The sampling module is used to acquire pipeline corrosion products in several pipelines.

[0156] The screening module obtains microscopic characterization data of pipeline corrosion products, standardizes the microscopic characterization data to obtain microscopic standard data, and screens the microscopic standard data based on the suitability test of factor analysis to obtain microscopic validation data.

[0157] The analysis module performs dimensionality reduction analysis on the micro-validation data, extracts common factors of several targets, and constructs a factor loading matrix. The factor loading matrix includes the loadings on the common factors of several targets. The common factors are divided into positive factors and negative factors based on the preset factor direction.

[0158] The output module couples the factor loading matrix with the positive and negative factors and calculates the comprehensive evaluation data of the pipeline corrosion status. If the comprehensive evaluation data is less than the preset value, the output pipeline is at risk of corrosion; otherwise, the output pipeline is in a healthy state.

[0159] Example 1

[0160] This embodiment takes a local water supply network as the implementation background. Based on the method established in the embodiment, the commonly used principal component analysis method is used for dimensionality reduction. The conditions for the validity of factor analysis include KMO statistic > 0.5 and Bartlett test p value < 0.05.

[0161] First, the raw data for the microscopic characterization of corrosion products on the inner wall of the pipeline were standardized:

[0162]

[0163] Calculate the correlation matrix or covariance matrix of the variables:

[0164]

[0165] To test the suitability of factor analysis, the Kaiser-Meyer-Olkin test (KMO test) is used to assess whether the correlation between variables is suitable for factor analysis. A KMO > 0.5 indicates that factor analysis is suitable.

[0166]

[0167] Bartlett's test of sphericity is used to determine whether the correlation matrix is ​​an identity matrix (no correlation). If the p-value is <0.05, it indicates that factor analysis is suitable.

[0168]

[0169] After screening the indicator data that can pass the test, factors are extracted using the eigenvalue decomposition method, and eigenvalue decomposition is performed on the correlation matrix R:

[0170] R = P∧P T

[0171] After calculation, the KMO statistic for this data set was 0.563, and the Bartlett's test of sphericity showed p < 0.001, meeting the requirements for factor analysis. After principal component analysis, the eigenvalues ​​of the first three principal components were greater than 1 (as shown in the scree plot). Figure 2 As shown, the component number represents the component in the microscopic characterization data. Component number 1 represents the first principal component (iron), component number 2 represents the second principal component (average pore size), and so on. The cumulative variance contribution rate of the first three principal components is 83.6%, which is greater than 80%, indicating that the first three principal components can be extracted to describe the level of pipeline corrosion.

[0172] The principal component coefficient matrix is ​​obtained by eigenvalue decomposition based on the extracted common factors, and the results are shown in Table 1.

[0173] Corrosion Characteristic Indicators PC1 PC2 PC3 Fe -0.843 0.208 -0.472 Mg 0.738 -0.61 -0.036 P 0.241 -0.433 0.829 Si 0.829 -0.054 0.385 Specific surface area 0.236 -0.827 0.193 S -0.11 0.743 -0.222 Average aperture 0.328 0.804 0.113 Ca 0.864 0.164 -0.156 Mn 0.023 0.036 0.952

[0174] Table 1

[0175] Among the principal components, Fe had the largest positive loading in PC1, indicating that this component is a positive factor, and the low fraction in PC1 reflects a large amount of corrosion products. In PC2, the average pore size had the largest positive loading, indicating that this component is a negative factor, and the high fraction in PC2 reflects high porosity and easy sediment detachment. In PC3, Mn had the largest positive loading, indicating that this component is a negative factor, and the high fraction in PC3 reflects strong manganese deposition in the aquatic environment. Overall, the principal components extracted from this set of data can reveal the corrosion state characteristics of the pipeline from corrosion and mineralization properties, sediment structural porosity, and environmental chemical characteristics.

[0176] Based on the extracted common factors, the specific value of each sample on the extracted factors is calculated using regression, resulting in the factor score matrix F:

[0177] F = R -1 L(L T R -1 L) -1 X

[0178] The elemental distribution and microstructure characterization of 18 groups of pipeline corrosion product samples were weighted according to the variance contribution rate of each common factor to calculate the comprehensive evaluation score of the corrosion state for each pipeline. This score was then validated using Spearman correlation analysis with subjective pipeline health scores based on actual images of the pipeline interior. Figure 3 As shown, the correlation coefficient is 0.91, indicating that the calculated score can well represent the corrosion status of the pipeline. The subjective pipeline health score is based on the corrosion level score of the pipeline photographs.

[0179] Comparative Example 1

[0180] A method of directly weighting and summing corrosion characterization indicators was selected as a comparative example. Keeping other application conditions the same as in the previous examples, the pipeline corrosion product characterization data in Example 1 were weighted according to the following principles: among the main corrosion-related elements (Fe, Mn, S), iron is the main corrosion product, manganese can promote localized corrosion, and sulfur can form H2S under anaerobic conditions, accelerating corrosion; therefore, they have relatively high weights, all at 0.15. Deposition-related elements (Ca, Mg, Si, P) are usually related to mineral scaling and can affect the formation and stability of corrosion products, thus they are assigned medium weights of 0.06–0.10. Specific surface area (BET) and average pore size affect the structural characteristics of corrosion products, such as porosity and density, and are assigned relatively low weights of 0.05. The scores calculated by directly weighting the standardized data according to the above weights were verified by Spearman correlation analysis with the subjective pipeline health score based on actual pipeline images. Figure 4 As shown, the correlation coefficient is only 0.38, indicating that this weighting method is rather crude and difficult to accurately describe pipeline corrosion. The existing pipeline health assessment parameters are scores obtained by weighting expert ratings based on actual images of the pipeline interior.

[0181] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for extracting multi-dimensional features of pipeline corrosion products and evaluating corrosion status, characterized in that, include: The method involves obtaining corrosion products from several pipelines; microscopic characterization data includes elemental index data, pore volume data, average pore size data, specific surface area data, and crystal structure data of the corrosion products; the elemental index data includes the relative content data of iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese, and chlorine in the corrosion products; the elemental index data is obtained by at least one of X-ray fluorescence spectroscopy, X-ray photoelectron spectroscopy, inductively coupled plasma spectroscopy, and scanning electron microscopy-energy dispersive spectroscopy; the pore volume data, average pore size data, specific surface area data, and crystal structure data are obtained by at least one of fully automated specific surface area and porosity analysis, X-ray diffraction, and scanning electron microscopy. Microscopic characterization data of the pipeline corrosion products are obtained, and the microscopic characterization data are standardized to obtain microscopic standard data. Based on the suitability test of factor analysis, the microscopic standard data are screened to obtain microscopic verification data. The microscopic verification data is subjected to dimensionality reduction analysis, and common factors of several targets are extracted to construct a factor loading matrix. The factor loading matrix includes loadings on common factors of several targets. The common factors are divided into positive factors and negative factors based on a preset factor direction. Specifically: when the loading value of iron content in the first common factor is less than -0.5, and the loading values ​​of calcium and silicon content in the first common factor are greater than 0.5, the common factor corresponding to iron is a positive factor; when the loading value of pore size data in the second common factor is less than -0.5, the common factor corresponding to pore size data is a positive factor; when the loading values ​​of manganese and phosphorus content in the third common factor are less than -0.5, the common factors corresponding to manganese and phosphorus are positive factors; if the loading direction of the main contributing feature in the common factor is inconsistent with the preset factor direction, the corresponding common factor is divided into negative factors according to the sign pair corresponding to the loading value in the common factor. The factor loading matrix is ​​coupled with the positive and negative factors to calculate the comprehensive evaluation data of the pipeline corrosion status. If the comprehensive evaluation data is less than the preset value, the pipeline is output as having a corrosion risk; otherwise, the pipeline is output as being in a healthy state.

2. The method for multi-dimensional feature extraction and corrosion state evaluation of pipeline corrosion products according to claim 1, characterized in that: The elemental index data, pore volume data, average pore size data, specific surface area data, and crystal structure data are standardized to obtain microscopic standard data. The standardization process includes data transformation and data standardization; The data transformation includes performing a central logarithmic ratio transformation on the relative content data to eliminate the pseudo-correlation between iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese and chlorine in pipeline corrosion products caused by the constant total amount; The data standardization includes Z-score standardization of the micro-standard data, and the Z-score standardization formula is: ； Where x represents the original data for microscopic characterization. For: Standardized data, For: sample mean, Sample standard deviation.

3. The method for multi-dimensional feature extraction and corrosion state evaluation of pipeline corrosion products according to claim 2, characterized in that: The suitability test based on factor analysis filters the micro-standard data to obtain micro-validation data, including: The data correlation matrix is ​​obtained from the micro-standard data, and the micro-validation data is obtained by screening the micro-standard data based on the suitability test of factor analysis. A data matrix is ​​constructed from the standardized microscopic standard data, and this data matrix is ​​then transformed into a standardized data matrix. The standardized data matrix is... The data correlation matrix is ​​obtained based on the standardized data matrix; The data correlation matrix is ​​as follows: in For: Standardized data matrix Transpose For: sample size; The suitability tests for the factor analysis include the KMO test and the Bartlett test for sphericity, and the micro-standard data are screened to obtain micro-validation data based on the preset values ​​of the KMO test and the Bartlett test for sphericity.

4. The method for multi-dimensional feature extraction and corrosion state evaluation of pipeline corrosion products according to claim 3, characterized in that: The micro-validation data is subjected to dimensionality reduction analysis, and common factors of several targets are extracted and a factor loading matrix is ​​constructed. The factor loading matrix includes loadings on the common factors of several targets, including: The dimensionality reduction analysis method includes principal component analysis, which includes: Perform eigenvalue decomposition on the data correlation matrix R: ； in For: eigenvector matrix, For: the transpose of the eigenvector matrix, For: diagonal eigenvalue matrix, The eigenvalues ​​in the equation are the eigenvalues; Based on the criterion that the eigenvalues ​​are greater than 1, the common factors of the first few objectives are selected, and the number of common factors is K. A factor loading matrix L of the common factors of the first few objectives is then constructed. ； The proportion of explained variance of the i-th common factor among the K common factors: in Let be the eigenvalue corresponding to the i-th common factor, and k be the number of common factors extracted.

5. The method for multi-dimensional feature extraction and corrosion state evaluation of pipeline corrosion products according to claim 4, characterized in that: The common factors of the plurality of objectives include a first common factor, a second common factor, and a third common factor; Among them, the iron content is the dominant variable in the first common factor, and the loading signs of calcium and silicon are opposite to those of iron, so as to reflect the inverse relationship between metal corrosion and surface protective layer, and to characterize the degree of internal wall corrosion and the state of protective layer. The second common factor uses the pore size parameter as the dominant variable to reflect the porosity characteristics of the corrosion product structure and to indicate the compactness and integrity of the corrosion layer. The third common factor has manganese content as the dominant variable and is positively correlated with phosphorus content, reflecting the enrichment degree of exogenous elements in corrosion products and is used to evaluate the component migration characteristics during the corrosion process.

6. The method for multi-dimensional feature extraction and corrosion state evaluation of pipeline corrosion products according to claim 5, characterized in that: The comprehensive evaluation data for calculating the pipeline corrosion state by coupling the factor loading matrix with positive and negative factors includes: The factor score matrix F is calculated using regression analysis on the aforementioned common factor data. in For: data correlation matrix, For: factor loading matrix, For: Standardized data matrix: Based on the common factor terms of the factor score matrix F , , Coupled with the corresponding common factor weights, the comprehensive evaluation score Y of the pipeline corrosion state is calculated: Wherein, the common factor weight is the proportion of the variance explained by the common factor, and Y1, Y2 and Y3 represent the weighted contribution values ​​of the first common factor, the second common factor and the third common factor, respectively; If the overall evaluation score of the corrosion state is less than 0, the output pipeline is at risk of corrosion; otherwise, the output pipeline is in a healthy state.

7. The method for multi-dimensional feature extraction and corrosion state evaluation of pipeline corrosion products according to claim 6, characterized in that: Also includes: The overall evaluation score of the corrosion state is compared with the existing pipeline health evaluation parameters using Spearman analysis or Pearson analysis to obtain the correlation coefficient data. If the correlation coefficient data is higher than 0.9, it indicates that the dimensionality reduction analysis is effective. The existing pipeline health evaluation parameters are obtained by scoring the pipeline health based on four dimensions: actual images of the pipeline interior, the proportion of rust area, the thickness of corrosion nodules, the surface smoothness, and the integrity of the pipe wall.

8. A system for extracting multi-dimensional features of pipeline corrosion products and evaluating corrosion status, characterized in that, The system is applied to the method of any one of claims 1-7: The sampling module is used to acquire pipeline corrosion products in several pipelines. Microscopic characterization data includes elemental index data, pore volume data, average pore size data, specific surface area data, and crystal structure data of the pipeline corrosion products. The elemental index data includes the relative content data of iron, calcium, magnesium, aluminum, silicon, phosphorus, sulfur, manganese, and chlorine in the pipeline corrosion products. The elemental index data is acquired using at least one of X-ray fluorescence spectroscopy, X-ray photoelectron spectroscopy, inductively coupled plasma spectroscopy, and scanning electron microscopy-energy dispersive spectroscopy. The pore volume data, average pore size data, specific surface area data, and crystal structure data are acquired using at least one of fully automated specific surface area and porosity analysis, X-ray diffraction, and scanning electron microscopy. The screening module acquires the microscopic characterization data of the pipeline corrosion products, standardizes the microscopic characterization data to obtain microscopic standard data, and screens the microscopic standard data based on the suitability test of factor analysis to obtain microscopic verification data. The analysis module performs dimensionality reduction analysis on the microscopic verification data, extracts common factors of several targets, and constructs a factor loading matrix. The factor loading matrix includes loadings on the common factors of several targets. Based on a preset factor direction, the common factors are divided into positive and negative factors: For example, when the loading value of iron content in the first common factor is less than -0.5, and the loading values ​​of calcium and silicon content in the first common factor are greater than 0.5, the common factor corresponding to iron is a positive factor; when the loading value of pore size data in the second common factor is less than -0.5, the common factor corresponding to pore size data is a positive factor; when the loading values ​​of manganese and phosphorus content in the third common factor are less than -0.5, the common factors corresponding to manganese and phosphorus are positive factors. If the loading direction of the main contributing feature in the common factor is inconsistent with the preset factor direction, the corresponding common factor is classified as a negative factor according to the sign pair corresponding to the loading values ​​in the common factor. The output module couples the factor loading matrix with the positive and negative factors and calculates the comprehensive evaluation data of the pipeline corrosion status. If the comprehensive evaluation data is less than the preset value, the output pipeline is considered to be at risk of corrosion; otherwise, the output pipeline is considered to be in a healthy state.

Images