Method for evaluating surface corrosion degree of metal material

Abstract

The invention provides a metal material surface corrosion degree evaluation method, which comprises the following steps: carrying out scanning measurement on a carbon steel surface by adopting an electrochemical impedance spectrometer to obtain impedance data under different frequencies, drawing a Nyquist diagram and a Bode diagram according to the impedance data, and fitting through an equivalent circuit model to obtain a resistance value and a capacitance value of the carbon steel surface; according to the obtained surface resistance value and capacitance value, in combination with the element composition and chemical state of the pitting area and the thickness and component information of the oxidation film, establishing a mapping relation between the resistance value and the pitting morphology characteristic and a mapping relation between the resistance value and the oxidation film characteristic; judging whether the element composition and the chemical state of the pitting corrosion area are obviously changed or not, and judging whether the thickness and component information of the oxidation film needs to be re-acquired and the corresponding relation between the resistance value and the characteristics of the oxidation film needs to be updated or not by combining the mapping relation between the resistance value and the characteristics of the oxidation film.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for evaluating the degree of surface corrosion of a metal material. Background Art

[0002] The assessment of pitting corrosion on carbon steel surfaces presents a technical challenge: oxide films and corrosion products form in the pitting areas, altering the conductive properties of the material surface. Traditional visual inspection methods struggle to accurately quantify the extent of corrosion and accurately reflect the initiation and progression of pitting in real time. Electrochemical measurements are susceptible to interference from environmental factors. Accurately assessing the depth and extent of pitting without damaging the material surface has become an urgent challenge. In particular, in practical engineering applications, how to quickly and accurately complete measurements and data analysis under complex environmental conditions to achieve real-time monitoring and assessment of the extent of pitting is a technical bottleneck that urgently needs to be overcome. Summary of the Invention

[0003] The present invention provides a method for evaluating the degree of surface corrosion of metal materials, which mainly includes:

[0004] An electrochemical impedance spectrometer was used to scan and measure the carbon steel surface, obtaining impedance data at different frequencies. Nyquist and Bode plots were drawn based on the impedance data, and the surface resistance and capacitance of the carbon steel were obtained by fitting an equivalent circuit model.

[0005] Based on the acquired surface resistance and capacitance values, combined with the elemental composition, chemical state, and oxide film thickness and composition information of the pitting area, a mapping relationship between resistance value and pitting morphology characteristics, and a mapping relationship between resistance value and oxide film characteristics are established;

[0006] Based on the mapping relationship between resistance value and pitting morphology characteristics, the corrosion potential and corrosion current density of carbon steel are obtained through electrochemical polarization curve testing. The corrosion rate is calculated by combining the Tafel extrapolation method, and a correlation model between corrosion rate and resistance value changes is established to achieve quantitative assessment of corrosion degree based on resistance value.

[0007] Based on the correlation model between corrosion rate and resistance value change, electrochemical noise measurement is used to obtain current noise and potential noise data on the carbon steel surface. The noise signal is analyzed in time and frequency by wavelet transform, and characteristic parameters are extracted. Combined with the resistance value change trend, an electrical characteristic model of the pitting initiation and expansion process is established.

[0008] Determine whether the elemental composition and chemical state of the pitting area have changed significantly. Combined with the mapping relationship between resistance value and oxide film characteristics, determine whether it is necessary to re-acquire the oxide film thickness and composition information and update the corresponding relationship between resistance value and oxide film characteristics.

[0009] Based on the acquired carbon steel surface resistance, the correlation model between corrosion rate and resistance change, the electrical characteristic model of pitting initiation and expansion, and the updated correspondence between resistance and oxide film characteristics, a comprehensive evaluation model was constructed using the support vector machine algorithm. Nonlinear classification was achieved through kernel function mapping to obtain quantitative evaluation results of the degree of pitting corrosion on the carbon steel surface.

[0010] If the evaluation results show that the pitting corrosion degree exceeds the preset threshold, protective measures need to be taken in a timely manner, and the relevant parameters in the correlation model between corrosion rate and resistance value change need to be adjusted to improve the accuracy of the evaluation and realize the pitting corrosion degree evaluation of the carbon steel surface of the metal material.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The present invention discloses a method for evaluating the degree of surface corrosion of metal materials. The method obtains impedance data of the carbon steel surface through electrochemical impedance spectroscopy measurement, and establishes a mapping relationship between resistance value and pitting morphology characteristics and oxide film characteristics. Combined with electrochemical polarization curve and electrochemical noise measurement, a correlation model between corrosion rate and resistance value change and an electrical characteristic model of pitting initiation and expansion process are constructed. By monitoring the changes in elemental composition and chemical state of the pitting area, the correspondence between resistance value and oxide film characteristics is dynamically updated. Finally, a support vector machine algorithm is used to integrate the above models and relationships to achieve quantitative evaluation of the degree of pitting corrosion on the carbon steel surface. The present invention can accurately evaluate the degree of pitting corrosion on the carbon steel surface, promptly detect situations that exceed the threshold, provide a basis for taking protective measures and optimizing evaluation models, and effectively improve the service life and safety of metal materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flow chart of a method for evaluating the degree of surface corrosion of a metal material.

[0014] Figure 2 Schematic diagram of a method for evaluating the degree of surface corrosion of a metal material according to the present invention. DETAILED DESCRIPTION

[0015] To further understand the content of the present invention, the present invention is described in detail with reference to the accompanying drawings and examples. The present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are intended only to illustrate the relevant invention and are not intended to limit the invention. It should also be noted that, for ease of description, only portions relevant to the invention are shown in the accompanying drawings.

[0016] like Figure 1 -2. In this embodiment, a method for evaluating the degree of corrosion of a metal material surface may specifically include:

[0017] In step S101, an electrochemical impedance spectrometer is used to scan and measure the surface of carbon steel to obtain impedance data at different frequencies, and a Nyquist plot and a Bode plot are drawn based on the impedance data. The surface resistance and capacitance of the carbon steel are obtained by fitting an equivalent circuit model.

[0018] The impedance data obtained by scanning and measuring the surface of the carbon steel sample by an electrochemical impedance spectrometer within a preset frequency range is obtained. The impedance data includes the impedance amplitude and phase angle information at different frequencies. The impedance data is processed using the FastFourierTransform algorithm to obtain the Nyquist plot and the Bode plot. Based on the Nyquist plot and the Bode plot, the impedance amplitude and phase angle parameters are extracted by fitting the curve using the least squares method. From the pre-established equivalent circuit model library, the best fitting model is selected through traversal comparison. If the fitting error is less than the preset threshold, the equivalent circuit model is determined to be the optimal model. Based on the optimal equivalent circuit model, the resistance and capacitance values ​​of the carbon steel surface are calculated using the circuit element parameters.

[0019] Specifically, an electrochemical impedance spectrometer was used to scan and measure the surface of a carbon steel sample within a frequency range of 10 Hz to 100 kHz. Impedance data of the carbon steel surface at different frequencies were obtained and stored in a database for subsequent analysis and processing. The stored impedance data were processed using the Fast Fourier Transform algorithm to generate Nyquist and Bode plots. The impedance amplitude and phase angle parameters were extracted and recorded in a database using the least squares curve fitting method. The best-fitting model was selected from a pre-established model library through traversal comparison. Parameter fitting was performed using the Levenberg-Marquardt nonlinear least squares algorithm, with a fitting error threshold of 0.01. If the fitting error was less than the preset threshold, the equivalent circuit model was determined to be the optimal model. Based on the optimal equivalent circuit model, the resistance and capacitance values ​​of the carbon steel surface were calculated using the circuit component parameters. The obtained parameter values ​​were stored in a database for subsequent analysis of the carbon steel surface characteristics. The electrochemical impedance spectrometer (EIS) scanned the surface of a carbon steel sample over a frequency range of 10 Hz to 100 kHz, with 100 frequency points. Ten measurements were taken at each frequency point and the average was calculated to enhance data reliability. The measured impedance data, including both real and imaginary components, were stored in a relational database, with one record for each frequency point. The Fast Fourier Transform algorithm was used to process the impedance data, converting the time-domain data into the frequency-domain data to generate Nyquist and Bode plots. The Nyquist plot displays the real impedance component on the abscissa and the negative imaginary impedance component on the ordinate. The Bode plot contains two subplots: one showing the impedance modulus as a function of frequency, and the other showing the phase angle as a function of frequency. Least squares fitting was performed on the Nyquist plot to extract characteristic parameters such as solution resistance Rs, charge transfer resistance Rct, and double-layer capacitance Cdl. Appropriate models were selected from a library of pre-established equivalent circuit models, such as the simple Randles circuit or the Warburg impedance model that accounts for diffusion effects. Parameter fitting was performed using the Levenberg-Marquardt nonlinear least-squares algorithm, with a maximum number of iterations set to 1000 and a convergence threshold of 0.01. If the fitting error was less than the threshold, the model was determined to be optimal. Based on the optimal equivalent circuit model, the resistance and capacitance values ​​of the carbon steel surface were calculated, for example, a charge transfer resistance (Rct) value of 1000 Ω·cm² and a double-layer capacitance (Cdl) value of 20 μF / cm². These parameter values ​​were stored in a database for subsequent analysis of the corrosion resistance and electrochemical properties of the carbon steel surface.

[0020] Step S102 , based on the acquired surface resistance and capacitance values, combined with the elemental composition, chemical state, and oxide film thickness and composition information of the pitting area, establish a mapping relationship between resistance and pitting morphology characteristics, and a mapping relationship between resistance and oxide film properties.

[0021] Obtain the resistance and capacitance values ​​of the carbon steel surface; scan the carbon steel surface with an optical microscope and identify the position coordinates of the pitting area through image analysis; measure the elemental composition and chemical state of the pitting area with an X-ray photoelectron spectrometer; determine the thickness of the oxide film with an ellipsometer; analyze the composition of the oxide film with an Auger electron spectrometer; store the obtained data in a feature database; collect a high-resolution image of the pitting area with a scanning electron microscope; process the high-resolution image with ImageJ software; extract pitting morphology characteristic parameters using threshold segmentation and edge detection methods; and associate the pitting morphology characteristic parameters with the resistance value in the feature database. ; A nonlinear regression model is used to establish a mapping relationship between the resistance value and the characteristic parameters of the pitting morphology; the mean square error is minimized by the gradient descent method to optimize the nonlinear regression model parameters; based on the resistance value, capacitance value, oxide film thickness and composition information in the characteristic database, the principal component analysis method is used to reduce the data dimension; a support vector machine regression algorithm is used to construct a two-dimensional input mapping relationship between the resistance value and capacitance value and the oxide film characteristics; the performance of the linear kernel, polynomial kernel and radial basis function kernel are compared; the kernel function with the smallest mean square error is selected; the grid search method is used to determine the optimal parameters of the kernel function; and the accuracy of the prediction of the oxide film characteristics is improved through cross-validation.

[0022] Specifically, the resistance and capacitance values ​​of the carbon steel surface are obtained from a database. The carbon steel surface is scanned using an optical microscope, and the coordinates of the pitting areas are identified through image analysis. For the identified pitting areas, the elemental composition and chemical state are measured using an X-ray photoelectron spectrometer, the oxide film thickness is determined using an ellipsometer, and the oxide film composition is analyzed using an Auger electron spectrometer. The resulting data is then stored in a feature database. High-resolution images of the pitting areas are collected using a scanning electron microscope. The images are processed using ImageJ software, and threshold segmentation and edge detection methods are applied to extract the pitting morphological characteristic parameters, including pitting depth, diameter, and density. The extracted morphological parameters are then correlated with the resistance values ​​in the feature database. A nonlinear regression model, such as polynomial regression or exponential regression, is used to establish a mapping relationship between the resistance value and the pitting morphological characteristic parameters. The gradient descent method is used to minimize the mean square error, optimize the model parameters, and obtain a mathematical expression for the resistance value and the pitting depth, diameter, and density. Based on the resistance, capacitance, oxide film thickness, and composition information in the feature database, principal component analysis was used to reduce data dimensionality. A support vector machine regression algorithm was then used to construct a two-dimensional input mapping between resistance and capacitance values ​​and oxide film properties. The performance of linear, polynomial, and radial basis function kernels was compared, and the kernel function with the lowest mean square error was selected. A grid search method was used to determine the optimal kernel parameters, and cross-validation was used to improve the accuracy of oxide film property prediction. During carbon steel surface analysis, resistance and capacitance values ​​were first extracted from the database. An optical microscope scanned a 1 cm² area of ​​the carbon steel surface at 50x magnification. Image analysis software used a grayscale thresholding method to identify pitting areas and record their coordinates. An X-ray photoelectron spectrometer scanned the pitting areas with an energy range of 0–1100 eV, with a step size of 0.1 eV, to obtain peak intensities and binding energies for Fe, Cr, and Ni. Ellipsometers were used to measure oxide film thickness at a 70° angle of incidence, obtaining values ​​such as 20 nm. Auger electron spectroscopy was used to analyze oxide film composition with an electron beam energy of 3 keV, generating elemental depth distribution curves. Images of the pitting areas were acquired using a scanning electron microscope at a magnification of 10,000x with a resolution of 4096 × 3072 pixels. ImageJ software was used to segment the images using the Otsu thresholding method, and the Canny edge detection algorithm was used to extract the pitting outlines. The pitting depth, diameter, and density were measured. A cubic polynomial regression model was used to establish the relationship between resistance and morphological parameters, and the mean square error was minimized using gradient descent. Dimensionality reduction was performed using principal component analysis, retaining the principal components that explained 95% of the variance. A support vector machine regression algorithm, using a radial basis function kernel, was used. Five-fold cross-validation and grid search were used to determine the optimal parameters, resulting in an R² value of 90% on the test set.

[0023] In step S103, based on the mapping relationship between resistance value and pitting morphology characteristics, the corrosion potential and corrosion current density of carbon steel are obtained through electrochemical polarization curve testing, the corrosion rate is calculated by combining the Tafel extrapolation method, and a correlation model between the corrosion rate and the resistance value change is established to achieve quantitative evaluation of the corrosion degree based on the resistance value.

[0024] According to the pre-established mapping relationship between resistance value and pitting morphology characteristics, the resistance value data of the carbon steel sample is obtained from the database; the polarization curve test is performed through the electrochemical workstation to obtain the corrosion potential and corrosion current density of the carbon steel; the polarization curve data is smoothed using the Savitzky-Golay filtering algorithm to eliminate noise; if the smoothing process is completed, the smoothed polarization curve data is processed using the Tafel extrapolation method; the linear region is determined by calculating the first-order derivative of the polarization curve; the corrosion current density is calculated using the slope intersection of the linear region, and the corrosion rate is converted according to Faraday's law; the resistance value time series data is processed using the exponential smoothing method; the resistance value change rate at adjacent time points is calculated, and the moving average method is used to obtain a smoothed change rate curve; the resistance value change rate, pitting depth, and pitting density are used as input features, and the principal component analysis method is used for dimensionality reduction; the support vector regression algorithm with a radial basis function kernel is used to establish a correlation model between the corrosion rate and the input features; the model parameters are optimized through cross-validation, and the optimal kernel function parameters are determined using a grid search method to obtain a quantitative assessment model of the corrosion degree.

[0025] Specifically, based on a pre-established mapping relationship between resistance and pitting morphology, resistance data for carbon steel samples were extracted from a database. Polarization curves were then measured using an electrochemical workstation with a scan rate of 0.5 mV / s and a potential range of ±250 mV from the corrosion potential to obtain the corrosion potential and corrosion current density. The acquired polarization curve data were smoothed and noise removed using a Savitzky-Golay filter algorithm with a window width of 15 data points and a polynomial order of 3. The processed data were then stored in a database. The smoothed polarization curve data were processed using the Tafel extrapolation method. By calculating the first-order derivative of the polarization curve, the region with an absolute value greater than 0.1 V / decade was identified as the linear region. A linear fit was performed to determine the Tafel slope, and the corrosion current density was calculated using the slope intersection points. The corrosion rate was then converted according to Faraday's law. Exponential smoothing was used to process the resistance time series data with a smoothing coefficient of α = 0.3. The rate of change of resistance at adjacent time points was calculated, and a smoothed rate of change curve was obtained using a five-point moving average. The resistance change rate, pitting depth, and pitting density were used as inputs for dimensionality reduction using principal component analysis, retaining the principal components that explained 95% of the variance. A correlation model between corrosion rate and multidimensional input features was established using a support vector regression algorithm with a radial basis function kernel. Model parameters were optimized through cross-validation, and the optimal kernel function parameters were determined using a grid search method, resulting in a quantitative assessment model for corrosion severity. For carbon steel corrosion assessment, resistance values ​​and pitting morphology features were first extracted from a database. Polarization curves were measured using an electrochemical workstation at a scan rate of 0.5 mV / s and a potential range of -250 mV to +250 mV. The measured polarization curve data were smoothed using Savitzky-Golay filtering. In Tafel extrapolation analysis, the linear region was identified by first-order derivative calculation, with slopes greater than 0.1 V / decade. Linear fitting yielded an anodic slope of βa = 60 mV / decade and a cathodic slope of βc = 120 mV / decade. The current density at the intersection of the slopes is the corrosion current density (icorr = 10 μA / cm²), which translates to a corrosion rate of 0.12 mm / year. The resistance time series was exponentially smoothed to calculate the rate of change, and then a 5-point moving average was applied to obtain a smoothed rate of change curve. Principal component analysis reduced the dimensionality of the input features, retaining two principal components. Support vector regression employed a radial basis function kernel. Parameters were optimized through 5-fold cross-validation and grid search. The final model achieved a goodness of fit of R² = 0.85 on the test set, enabling quantitative assessment of corrosion severity.

[0026] In step S104, based on the correlation model between corrosion rate and resistance value change, the current noise and potential noise data of the carbon steel surface are obtained by electrochemical noise measurement, the noise signal is subjected to time-frequency analysis by wavelet transform, characteristic parameters are extracted, and the electrical characteristic model of the pitting initiation and expansion process is established in combination with the resistance value change trend.

[0027] A pre-established model correlating corrosion rate and resistance change is obtained, and resistance change data for carbon steel samples is extracted from the correlation model. Current noise and potential noise data from the carbon steel surface are collected using an electrochemical noise measurement device. The collected noise data is preprocessed, and if a trend is present, detrending is performed. Median filtering is used to remove outliers. The filtered data is then normalized to zero mean to obtain preprocessed noise data. Discrete wavelet transform is used to perform time-frequency analysis on the preprocessed noise data. The number of decomposition levels is determined based on the frequency characteristics of the signal. Wavelet coefficients for different frequency bands are calculated. Feature parameters are extracted from the discrete wavelet transform results, including energy ratios for each frequency band, noise power spectral density, and noise resistance. Principal component analysis is used to reduce the dimensionality of the extracted feature parameters. The principal component whose cumulative contribution rate reaches a preset threshold is determined, and the noise feature parameters after dimensionality reduction are obtained. A long short-term memory neural network is constructed based on the resistance value change trend and the noise characteristic parameters after dimensionality reduction. An electrical characteristic model of the pitting initiation and expansion process is established through the neural network. If the pitting probability output by the model exceeds a preset threshold, it is determined that pitting has started.

[0028] Specifically, based on a pre-established model correlating corrosion rate with resistance change, resistance change data for carbon steel samples were extracted from a database. Current and potential noise data on the carbon steel surface were collected using an electrochemical noise measurement device, with a sampling frequency of 1 kHz and a sampling duration of 1800 seconds. The collected noise data were detrended, outliers were removed using a median filter, and then zero-mean normalized. The preprocessed noise data were then stored in a database. Discrete wavelet transforms were used to perform time-frequency analysis on the preprocessed current and potential noise data. The reconstruction errors of the wavelet basis functions db4, sym4, and coif4 were compared, and the wavelet basis function with the smallest reconstruction error was selected. The number of decomposition levels was determined based on the signal's frequency characteristics, ensuring that the lowest frequency band contained the pitting characteristic frequency. Wavelet coefficients were then obtained for each frequency band, and the energy distribution and standard deviation of each band were calculated. Feature parameters were extracted from the wavelet transform results, including the energy ratio of each frequency band, the noise power spectral density, and the noise resistance. The noise power spectral density was calculated using the Welch method, and the variance was reduced using the piecewise averaging method. The noise resistance was calculated by dividing the standard deviation of the potential noise by the standard deviation of the current noise. Principal component analysis was used to reduce the dimensionality of the extracted feature parameters, retaining the principal components with a cumulative contribution rate of 95%. Combining the resistance trend and the noise feature parameters after dimensionality reduction, a long short-term memory neural network was used to construct an electrical characteristic model for the pitting initiation and propagation process. This network consists of an input layer, two LSTM hidden layers, and a fully connected output layer. The Adam optimizer was used, and the loss function was the mean squared error. The number of training rounds was set to 500, and the learning rate was 0.001. The model parameters were optimized using the backpropagation algorithm. A threshold for pitting initiation was set; pitting initiation was determined when the pitting probability output by the model exceeded the threshold. The pitting propagation process was determined by the rate of change of the pitting probability within a continuous time window. For the carbon steel pitting analysis, resistance change data was first extracted from the database. Noise data was collected using an electrochemical noise measurement device with a sampling frequency of 1 kHz and a duration of 1800 seconds. The noise data was detrended, outliers were removed using a 5-point median filter, and then normalized to zero mean. The discrete wavelet transform uses the db4 wavelet basis function and 5-layer decomposition to obtain the wavelet coefficients of the frequency bands [0-15.625Hz, 15.625-31.25Hz, 31.25-62.5Hz, 62.5-125Hz, 125-250Hz]. Calculate the energy ratio of each frequency band

[0029] The power spectral density was calculated using the Welch method, with a noise resistance of 10Ω. Dimensionality reduction was performed using principal component analysis, with a cumulative contribution rate set to 95%, ultimately retaining three principal components. The long short-term memory neural network consisted of an input layer, two LSTM hidden layers, and a fully connected output layer. The Adam optimizer used a learning rate of 0.001, a batch size of 32, and training for 500 epochs. A pitting onset threshold of 0.7 was set, with the output probability exceeding 0.7 indicating pitting onset. A pitting expansion was identified when the rate of change in the pitting probability exceeded 0.05 / s within a continuous 10-second time window. The model achieved 90% accuracy on the test set, successfully identifying the pitting onset and expansion process.

[0030] Step S105 , determining whether the elemental composition and chemical state of the pitting area have changed significantly, and combining the mapping relationship between the resistance value and the oxide film characteristics, determining whether it is necessary to re-acquire the oxide film thickness and composition information and update the corresponding relationship between the resistance value and the oxide film characteristics.

[0031] Obtain data from an X-ray photoelectron spectrometer for elemental composition and chemical state analysis of the pitting area, including element peak intensity and binding energy information. Use principal component analysis to perform multivariate analysis on the data to obtain the element content change rate and the degree of chemical state change. If the element content change rate exceeds a preset threshold or the chemical state change degree exceeds a preset threshold, it is determined to be a significant change. If it is determined to be a significant change, a remeasurement instruction is sent to the data acquisition system. Receive the remeasurement instruction and start the oxide film characteristic remeasurement process. The oxide film thickness data measured by the ellipsometer and the oxide film composition data analyzed by the Auger electron spectrometer are obtained through the oxide film characteristic remeasurement process. Based on the oxide film thickness data and the oxide film composition data, a multi-layer model is used to fit the oxide film structure information. A support vector regression algorithm is used to establish a correspondence between resistance value and oxide film structure information, and the support vector regression algorithm selects a radial basis function kernel. The optimal parameters of the radial basis function kernel are determined by a grid search method to obtain a multidimensional mapping relationship model between resistance value and oxide film characteristics.

[0032] Specifically, an X-ray photoelectron spectrometer (XPS) was used to analyze the elemental composition and chemical state of the pitting area, with a scan range of 0-1100 eV and a step size of 0.1 eV. Element peak intensity and binding energy data were acquired. The acquired data were stored in a database and subjected to multivariate analysis using principal component analysis to calculate the element content change rate and the degree of chemical state change. Significance of the elemental composition and chemical state changes was determined based on pre-set thresholds. A significant change was determined if the element content change rate exceeded 10% or the chemical state change exceeded 5%. If a significant change was determined, a remeasurement instruction was sent to the data acquisition system, initiating the remeasurement process for the oxide film properties. The oxide film thickness was measured using an ellipsometer with an incident angle of 70 degrees and a wavelength range of 300-800 nm. The oxide film composition was analyzed using an Auger electron spectrometer (AES) with an electron beam energy of 3 keV, generating element depth distribution curves. The ellipsometer thickness data was fused with the element depth distribution curves acquired by the AES, and a multilayer model was used to fit the complete oxide film structure. A support vector regression algorithm was used to update the relationship between resistance and oxide film properties. A radial basis function kernel was selected, and kernel parameters were optimized using a grid search method. Model performance was evaluated using 5-fold cross-validation. The parameter combination with the lowest root mean square error (RMS) was selected as the optimal model. A multidimensional mapping relationship between resistance, oxide film thickness, and composition was established, accounting for nonlinear correlations to improve the model's prediction accuracy. The updated model was stored in a database. X-ray photoelectron spectroscopy was used to analyze the pitting area, scanning from 0 to 1100 eV with a step size of 0.1 eV. Peak intensities and binding energies of Fe2p, Cr2p, and O1s were obtained. Principal component analysis was used to process the data, revealing that the first principal component explained 85% of the variance. The calculated change in Fe content was 12%, and the change in Cr chemical state was 6%. Significant changes were identified, triggering a re-measurement command. Ellipsometer measurements of oxide film thickness were performed using an incident angle of 70 degrees and a wavelength of 300 to 800 nm, yielding a film thickness of 23 nm. Composition analysis was performed using an Auger electron spectrometer with an electron beam energy of 3 keV, generating depth distribution curves for Fe, Cr, and O. A multi-layer model was fitted to the fused data to determine the oxide film structure: an outer layer of Fe2O3, a middle layer of Fe3O4, and an inner layer of FeCr2O4. A support vector regression algorithm was used to update the mapping, and grid search was used to optimize the parameters. A 5-fold cross-validation root mean square error (RMS) of 0.05 was achieved. A multidimensional mapping between resistance and oxide film properties was established: R = f(d, CFe, CCr, CO), where R is the resistance, d is the thickness, and C is the element content. A polynomial kernel function with degree = 3 was used to account for nonlinear correlations. The updated model achieved an R2 of 0.92 on the test set, improving prediction accuracy by 15%. The model parameters and structure were stored in a database for subsequent corrosion monitoring and prediction.

[0033] In step S106, a comprehensive evaluation model is constructed using a support vector machine algorithm based on the acquired carbon steel surface resistance value, the correlation model between corrosion rate and resistance value change, the electrical characteristic model of the pitting initiation and expansion process, and the updated correspondence between resistance value and oxide film characteristics. Nonlinear classification is achieved through kernel function mapping to obtain a quantitative evaluation result of the degree of pitting corrosion on the carbon steel surface.

[0034] Obtain the data on the relationship between the surface resistance value of carbon steel, the correlation model parameters of the corrosion rate and the change of the resistance value, the pitting characteristic model parameters, and the corresponding relationship between the resistance value and the oxide film characteristics; interpolate the missing data using the moving average method based on the data, and identify the outliers using the 3σ criterion; map the eigenvalues ​​to the interval [0, 1] using the minimum-maximum scaling method; construct an eigenvector matrix containing the normalized resistance value, the corrosion rate prediction value, the pitting characteristic parameters, and the oxide film characteristic parameters; calculate the covariance matrix of the eigenvector matrix, solve the eigenvalues ​​and eigenvectors; select the top k eigenvalues ​​with a cumulative contribution rate of 95% The method comprises the following steps: a principal component is obtained, and dimensionality reduction processing is performed on the eigenvector; the dimensionality-reduced data is divided into a training set and a test set; a support vector machine algorithm is used to construct a comprehensive evaluation model, and a radial basis function is selected as the kernel function; the kernel function parameters and penalty coefficient are optimized by a grid search method, the parameter search range is set to C and γ, and the model performance is evaluated by cross-validation; the support vector machine model is used to perform nonlinear classification on the test data; according to the distance from the support vector to the separating hyperplane, the sigmoid function is used to map the distance to the interval [0, 1]; a quantitative evaluation value of the pitting corrosion degree is calculated, and the evaluation result is stored in a database.

[0035] Specifically, data on the surface resistance of carbon steel, model parameters for the correlation between corrosion rate and resistance change, parameters for the electrical characteristic model of pitting initiation and propagation, and updated data on the correspondence between resistance and oxide film properties were extracted from the database. Missing data were interpolated using the moving average method, and outliers were identified and processed using the 3σ criterion. All eigenvalues ​​were normalized by mapping them to the interval [0, 1] using the min-max scaling method. An eigenvector matrix was constructed, including the normalized resistance values, corrosion rate predictions, pitting characteristic parameters, and oxide film characteristic parameters. The characteristic covariance matrix was calculated, and the eigenvalues ​​and eigenvectors were solved. The top k principal components with a cumulative contribution rate of 95% were selected for dimensionality reduction. 80% of the reduced data was used as the training set, and 20% as the test set. A comprehensive evaluation model was constructed using the support vector machine algorithm, with the radial basis function as the kernel function. The kernel technique was used to map low-dimensional features to a high-dimensional feature space. A linear classifier was constructed in this high-dimensional space, achieving nonlinear classification in the original space. The dual problem of the support vector machine is solved using the sequential minimum optimization algorithm, and the linear inseparable case is handled by maximizing the soft interval. The kernel function parameters and penalty coefficients are optimized by the grid search method, and the parameter search range is set to C: [0.1, 100], γ: [0.001, 1]. The model performance is evaluated using 5-fold cross validation. The trained support vector machine model is used to perform nonlinear classification on the test data. The pitting degree is divided into 5 levels: mild, light, moderate, severe, and extremely severe. According to the distance from the support vector to the separating hyperplane, the sigmoid function is used to map the distance to the interval [0, 1] and correspond to the corresponding level. The quantitative evaluation value of the pitting degree is calculated and the evaluation results are stored in the database. In the process of carbon steel pitting evaluation, the resistance value, corrosion rate, pitting characteristics and oxide film characteristic data are first extracted from the database. The 5-point moving average method is used to interpolate the missing data, and the 3σ criterion is used to eliminate outliers. The minimum-maximum scaling normalizes the eigenvalues ​​to

[0036] [0, 1]. Construct a 10-dimensional eigenvector, calculate the covariance matrix, and solve for the eigenvalues ​​and eigenvectors. Dimensionality reduction was performed by selecting the top six principal components with a cumulative contribution rate of 95%. The 1000 data sets were divided into 800 training sets and 200 test sets. The support vector machine model used the RBF kernel function and the sequential minimization algorithm to solve the dual problem, with a soft margin parameter of C = 10. A grid search was used to optimize the parameters C = [0.1, 100] and γ = [0.001, 1], achieving a minimum mean squared error of 0.05 for 5-fold cross-validation. Nonlinear classification mapped the samples into a high-dimensional space and constructed the optimal separating hyperplane. The severity of pitting corrosion was categorized into five levels: 0-0.2 (mild), 0.2-0.4 (mild), 0.4-0.6 (moderate), 0.6-0.8 (severe), and 0.8-1.0 (extremely severe). The distance from the support vector to the hyperplane was mapped to [0, 1] using a sigmoid function. For example, a distance of 2 corresponds to an evaluation value of 0.88. The model achieved an accuracy of 92% on the test set, and the evaluation results were stored in the database for subsequent analysis.

[0037] In step S107, if the evaluation result shows that the pitting corrosion degree exceeds the preset threshold, protective measures need to be taken in time, and the relevant parameters in the correlation model between the corrosion rate and the resistance value change need to be adjusted to improve the accuracy of the evaluation and realize the pitting corrosion degree evaluation of the carbon steel surface of the metal material.

[0038] Obtain the pitting corrosion degree assessment result and the preset threshold stored in the database, and compare the pitting corrosion degree assessment result with the preset threshold; if the pitting corrosion degree assessment result exceeds the preset threshold, trigger the protection measure execution instruction; start the cathodic protection system according to the protection measure execution instruction, adjust the protection current density of the cathodic protection system, and control the corrosion inhibitor injection device to adjust the corrosion inhibitor concentration and injection rate; use the sliding time window method to calculate the corrosion rate and resistance value change data to obtain updated corrosion rate data and resistance value change data; use the weighted least squares method to process the updated corrosion rate data and resistance value change data to obtain updated correlation model parameters; evaluate the updated correlation model parameters through the cross-validation method to determine whether the updated correlation model parameters meet the preset accuracy conditions. If the preset accuracy conditions are met, adopt the updated correlation model parameters; if the preset accuracy conditions are not met, retain the original correlation model parameters.

[0039] Specifically, the system retrieves the latest pitting assessment results and preset thresholds from a database. The preset pitting threshold is set at 0.7. A comparison algorithm determines whether the pitting level exceeds the preset threshold. If so, the protective action execution instructions and model parameter adjustment process are triggered. Based on the triggered protective action execution instructions, the cathodic protection system is automatically activated, adjusting the protection current density range to 0.1-1.0 mA / cm². Simultaneously, a signal is sent to the intelligent control system to control the corrosion inhibitor injection device, adjusting the inhibitor concentration range to 50-200 ppm and the injection rate to 0.1-0.5 L / h. A 30-day sliding time window method is used to recalculate the corrosion rate and resistance change data for the recent period, with data updated daily. The parameters of the correlation model between corrosion rate and resistance change are updated using the weighted least squares method, with recent data given a higher weight. The updated parameters are then stored in the database. Based on the updated correlation model parameters, the cathodic protection current density and corrosion inhibitor injection parameters are adjusted in real time, forming a closed-loop control system. The accuracy of the updated model was evaluated using a 5-fold cross-validation method, calculating the root mean square error and coefficient of determination (R²). If the new model's R² improved by more than 5%, the new model was adopted; otherwise, the original model was retained. The evaluation results and model status were recorded in a log file. In the carbon steel pitting corrosion protection system, the database recorded the pitting severity assessment results in real time. The latest value was 0.75, exceeding the preset threshold of 0.7. The system immediately triggered protective measures: the cathodic protection current density was adjusted from the initial 0.2 mA / cm² to 0.8 mA / cm², the corrosion inhibitor concentration was increased from 75 ppm to 150 ppm, and the injection rate was increased from 0.2 L / h to 0.4 L / h. Simultaneously, a 30-day sliding window data analysis was initiated, collecting daily corrosion rates and resistance values. The correlation model was updated using the weighted least squares method, with data weights of 0.6 for the last 7 days, 0.3 for days 8-14, and 0.1 for days 15-30. The new model function is CR = 0.0015 × R + 0.05, where CR is the corrosion rate and R is the resistance. Based on the new model, the closed-loop control system fine-tuned the cathodic protection current density to 0.75 mA / cm² and the corrosion inhibitor concentration to 140 ppm. A 5-fold cross-validation evaluation of the new model showed a reduction in root mean square error (RMS) from 0.015 to 0.012, and an increase in R² from 0.85 to 0.91, a 7.1% improvement. The system adopted the new model, recording the evaluation time as 2023-04-15 14:30, indicating a successful model update and a 7.1% performance improvement.

[0040] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for evaluating the degree of corrosion on the surface of a metal material. It is characterized in that The method comprises: The surface of carbon steel was scanned and measured using an electrochemical impedance spectrometer to obtain impedance data at different frequencies. Nyquist and Bode plots were drawn based on the impedance data, and the surface resistance and capacitance of carbon steel were obtained by fitting an equivalent circuit model. Based on the acquired surface resistance and capacitance values, combined with the element composition, chemical state, oxide film thickness and composition information of the pitting area, a mapping relationship between resistance value and pitting morphology characteristics, and a mapping relationship between resistance value and oxide film characteristics are established; Based on the mapping relationship between resistance value and pitting morphology, the corrosion potential and corrosion current density of carbon steel are obtained through electrochemical polarization curve test. The corrosion rate is calculated by combining Tafel extrapolation method, and a correlation model between corrosion rate and resistance value change is established to achieve quantitative evaluation of corrosion degree based on resistance value. Based on the correlation model between corrosion rate and resistance value change, the current noise and potential noise data of carbon steel surface are obtained by electrochemical noise measurement. The noise signal is analyzed by wavelet transform to extract characteristic parameters. Combined with the resistance value change trend, the electrical characteristic model of pitting initiation and expansion process is established. Determine whether the element composition and chemical state of the pitting area have changed significantly, and combine the mapping relationship between the resistance value and the oxide film characteristics to determine whether it is necessary to re-acquire the oxide film thickness and composition information and update the corresponding relationship between the resistance value and the oxide film characteristics; Based on the obtained carbon steel surface resistance value, the correlation model between corrosion rate and resistance value change, the electrical characteristic model of pitting initiation and expansion process, and the correspondence between the updated resistance value and oxide film characteristics, a comprehensive evaluation model was constructed using the support vector machine algorithm. Nonlinear classification was achieved through kernel function mapping to obtain the quantitative evaluation results of the pitting degree of carbon steel surface. If the evaluation results show that the degree of pitting corrosion exceeds the preset threshold, protective measures need to be taken in time and the relevant parameters in the correlation model between corrosion rate and resistance value change need to be adjusted to improve the accuracy of the evaluation and realize the evaluation of the degree of pitting corrosion on the surface of carbon steel of metal materials.

2. The method according to claim 1, in, The electrochemical impedance spectrometer is used to scan and measure the carbon steel surface, impedance data at different frequencies are obtained, Nyquist plots and Bode plots are drawn according to the impedance data, and the surface resistance and capacitance of the carbon steel are obtained by fitting an equivalent circuit model, including: Obtain impedance data obtained by scanning and measuring the surface of a carbon steel sample by an electrochemical impedance spectrometer within a preset frequency range, wherein the impedance data includes impedance amplitude and phase angle information at different frequencies; The impedance data is processed using the FastFourierTransform algorithm to obtain the Nyquist and Bode plots; According to the Nyquist diagram and Bode diagram, the impedance amplitude and phase angle parameters are extracted by fitting the curves using the least square method; From a pre-established equivalent circuit model library, the best fitting model is selected by traversal comparison, and if the fitting error is less than a preset threshold, the equivalent circuit model is determined to be the optimal model; According to the optimal equivalent circuit model, the resistance and capacitance values ​​of the carbon steel surface are calculated by using the circuit element parameters.

3. The method according to claim 1, in, The mapping relationship between the resistance value and the pitting morphology characteristics and the mapping relationship between the resistance value and the oxide film characteristics are established based on the acquired surface resistance value and capacitance value, combined with the element composition, chemical state, and oxide film thickness and component information of the pitting area, including: Obtain the resistance and capacitance values ​​of the carbon steel surface; The carbon steel surface was scanned using an optical microscope, and the position coordinates of the pitting area were identified through image analysis; For the pitting area, an X-ray photoelectron spectrometer is used to measure the element composition and chemical state; The oxide film thickness was measured by ellipsometer; The composition of the oxide film was analyzed using an Auger electron spectrometer; storing the obtained data in a feature database; collecting a high-resolution image of the pitting area using a scanning electron microscope; Process the high-resolution images using ImageJ software; Threshold segmentation and edge detection methods are used to extract pitting morphology feature parameters; Associating the pitting morphology characteristic parameters with the resistance values ​​in the characteristic database; A nonlinear regression model is used to establish a mapping relationship between the resistance value and the characteristic parameters of the pitting morphology; Optimizing the nonlinear regression model parameters by minimizing the mean square error through the gradient descent method; According to the resistance value, capacitance value, oxide film thickness and composition information in the feature database, principal component analysis is used to reduce the data dimension; The support vector machine regression algorithm is used to construct a two-dimensional input mapping relationship between resistance and capacitance values ​​and oxide film characteristics; Compare the performance of linear kernels, polynomial kernels, and radial basis function kernels; Select the kernel function with the smallest mean square error; Using a grid search method to determine the optimal parameters of the kernel function; Improve the accuracy of oxide film property prediction through cross validation.

4. The method according to claim 1, in, Based on the mapping relationship between the resistance value and the pitting morphology characteristics, the corrosion potential and corrosion current density of carbon steel are obtained by electrochemical polarization curve testing, the corrosion rate is calculated by combining the Tafel extrapolation method, and a correlation model between the corrosion rate and the resistance value change is established to achieve quantitative evaluation of the corrosion degree based on the resistance value, including: According to the pre-established mapping relationship between resistance value and pitting morphology characteristics, the resistance value data of the carbon steel sample is obtained from the database; Polarization curve test is carried out through electrochemical workstation to obtain the corrosion potential and corrosion current density of carbon steel; The polarization curve data is smoothed using the Savitzky-Golay filtering algorithm to eliminate noise. If the smoothing process is completed, the smoothed polarization curve data is processed using the Tafel extrapolation method; Determining a linear region by calculating a first-order derivative of the polarization curve; The corrosion current density is calculated using the slope intersection point of the linear region, and the corrosion rate is converted according to Faraday's law; The resistance value time series data is processed by using exponential smoothing method; Calculate the resistance value change rate at adjacent time points and use the moving average method to obtain a smooth change rate curve; The resistance value change rate, pitting depth and pitting density are used as input features, and the principal component analysis method is used for dimensionality reduction; The support vector regression algorithm with radial basis function kernel is used to establish the correlation model between corrosion rate and input features; The model parameters are optimized through cross-validation, and the optimal kernel function parameters are determined using a grid search method to obtain a quantitative assessment model for the degree of corrosion.

5. The method according to claim 1, in, The correlation model based on the corrosion rate and resistance value change uses electrochemical noise measurement to obtain current noise and potential noise data on the carbon steel surface, performs time-frequency analysis on the noise signal through wavelet transform, extracts characteristic parameters, and combines the resistance value change trend to establish an electrical characteristic model of the pitting initiation and expansion process, including: Obtain a pre-established correlation model between corrosion rate and resistance value change, and extract resistance value change data of the carbon steel sample from the correlation model; The electrochemical noise measurement device was used to collect the current noise and potential noise data on the carbon steel surface; Preprocess the collected noise data. If there is a trend in the noise data, perform detrending processing. Median filtering is used to remove outliers; Perform zero-mean normalization processing on the filtered data to obtain preprocessed noise data; For the pre-processed noise data, discrete wavelet transform is used to perform time-frequency analysis; Determine the number of decomposition layers according to the frequency characteristics of the signal; Calculate the wavelet coefficients of different frequency bands; Extract characteristic parameters from the discrete wavelet transform result, wherein the characteristic parameters include energy ratio of each frequency band, noise power spectrum density and noise resistance; The principal component analysis method is used to reduce the dimension of the extracted feature parameters; Determine the principal component whose cumulative contribution rate reaches a preset threshold; Obtain noise characteristic parameters after dimensionality reduction; Constructing a long short-term memory neural network according to the resistance value change trend and the noise characteristic parameters after the dimensionality reduction; An electrical characteristic model of the pitting initiation and expansion process is established by the neural network; If the pitting probability output by the model exceeds the preset threshold, it is determined that pitting has started.

6. The method according to claim 1, in, The determining whether the element composition and chemical state of the pitting area have changed significantly, combining the mapping relationship between the resistance value and the oxide film characteristics, determining whether it is necessary to re-acquire the oxide film thickness and composition information and update the corresponding relationship between the resistance value and the oxide film characteristics, includes: Obtain data from X-ray photoelectron spectrometer to analyze the elemental composition and chemical state of the pitting area, including element peak intensity and binding energy information; The principal component analysis method is used to perform multivariate analysis on the data to obtain the element content change rate and the degree of chemical state change; If the element content change rate exceeds a preset threshold or the chemical state change degree exceeds a preset threshold, it is determined to be a significant change; If it is determined to be a significant change, a re-measurement instruction is sent to the data acquisition system; receiving the re-measurement instruction and starting the oxide film characteristic re-measurement process; Obtain oxide film thickness data measured by ellipsometer and oxide film composition data analyzed by Auger electron spectrometer through oxide film property re-measurement process; According to the oxide film thickness data and oxide film composition data, the oxide film structure information is obtained by using a multi-layer model fitting; A support vector regression algorithm is used to establish a corresponding relationship between the resistance value and the oxide film structure information, and the support vector regression algorithm selects a radial basis function kernel; The optimal parameters of the radial basis function kernel are determined by a grid search method, and a multi-dimensional mapping relationship model between resistance value and oxide film characteristics is obtained.

7. The method according to claim 1, in, According to the obtained carbon steel surface resistance value, the correlation model between corrosion rate and resistance value change, the electrical characteristic model of pitting initiation and expansion process, and the correspondence between the updated resistance value and the oxide film characteristics, a support vector machine algorithm is used to construct a comprehensive evaluation model, and nonlinear classification is achieved through kernel function mapping to obtain a quantitative evaluation result of the degree of pitting corrosion on the carbon steel surface, including: Obtain the surface resistance value of carbon steel, the correlation model parameters of corrosion rate and resistance value change, the pitting characteristic model parameters, and the corresponding relationship data between resistance value and oxide film characteristics; Based on the data, the moving average method is used to interpolate the missing data, and the 3σ criterion is used to identify outliers; The minimum-maximum scaling method is used to map the eigenvalues ​​to the interval [0, 1]; Constructing a feature vector matrix including normalized resistance value, corrosion rate prediction value, pitting characteristic parameters and oxide film characteristic parameters; Calculating the covariance matrix of the eigenvector matrix, and solving the eigenvalues ​​and eigenvectors; Select the first k principal components with cumulative contribution rates of 95% and perform dimension reduction processing on the feature vector; Divide the reduced-dimensional data into training set and test set; The support vector machine algorithm is used to build a comprehensive evaluation model, and the radial basis function is selected as the kernel function; Optimize kernel function parameters and penalty coefficients through grid search, set parameter search range to C and γ, and use cross validation to evaluate model performance; Using the support vector machine model to perform nonlinear classification on the test data; According to the distance from the support vector to the separating hyperplane, the sigmoid function is used to map the distance to the interval [0, 1]; Calculate the quantitative evaluation value of the pitting degree and store the evaluation results in the database.

8. The method according to claim 1, in, If the evaluation result shows that the pitting degree exceeds the preset threshold, it is necessary to take protective measures in time and adjust the relevant parameters in the correlation model between the corrosion rate and the resistance value change to improve the accuracy of the evaluation and realize the pitting degree evaluation of the carbon steel surface of the metal material, including: Obtaining the pitting corrosion degree evaluation result and the preset threshold value stored in the database, and comparing the pitting corrosion degree evaluation result with the preset threshold value; If the pitting corrosion degree assessment result exceeds the preset threshold, a protective measure execution instruction is triggered; Starting the cathodic protection system according to the protection measure execution instruction, adjusting the protection current density of the cathodic protection system, and controlling the corrosion inhibitor injection device to adjust the corrosion inhibitor concentration and injection rate; The sliding time window method is used to calculate the corrosion rate and resistance value change data to obtain updated corrosion rate data and resistance value change data; Processing the updated corrosion rate data and resistance value change data using a weighted least squares method to obtain updated correlation model parameters; Evaluate the updated association model parameters by cross-validation method to determine whether the updated association model parameters meet the preset accuracy condition, and if so, adopt the updated association model parameters; If the preset accuracy condition is not met, the original correlation model parameters are retained.

Images