Coating damage evolution analysis method based on Raman spectrum

Abstract

The invention is applicable to the technical field of surface analysis and material characterization, and provides a coating damage evolution analysis method based on Raman spectrum, which comprises the following steps: step 1, detecting characteristic substances before coating damage; 2, detecting characteristic substances after the coating is damaged; 3, electrochemical corrosion performance testing before and after coating damage; 4, establishing a Raman spectrum database of the characteristic substances before and after damage; 5, training a coating damage evolution identification model; step 6, designing a coating damage evolution identification system; 7, verifying a system analysis result; according to the coating damage evolution analysis method based on the Raman spectrum provided by the invention, a Raman spectrum technology is taken as a core, technologies such as a scanning electron microscope / energy disperse spectroscopy and an electrochemical workstation are combined, and models such as a principal component analysis-support vector machine and a random forest are adopted for training comparison; and combining the model with the optimal performance with a spectrum preprocessing system to form an integrated coating damage evolution analysis system.

Description

Technical Field

[0001] The invention belongs to the technical field of surface analysis and material characterization, and in particular relates to a coating damage evolution analysis method based on Raman spectroscopy. Background Art

[0002] With the widespread application of functional coatings in industries such as machinery, aerospace, marine equipment, and electronic devices, various substrate materials often need to be coated with composite coatings of different proportions with properties such as corrosion resistance, wear resistance, insulation, or high temperature protection. The quality and reliability of the coating are directly related to the service life and safety of the product. Corrosion resistance is often one of the important indicators for evaluating the comprehensive performance of the coating. Therefore, it is crucial to have a deep understanding of the damage mechanism and evolution process of the coating under different working conditions for material research and development and engineering applications.

[0003] At present, the commonly used detection methods for the coating microstructure, components, and phase and chemical changes before and after damage include scanning electron microscopy, energy spectrum analysis, X-ray diffraction, and infrared spectroscopy. Although these methods have certain advantages in qualitative analysis and some semi-quantitative research, they usually require more cumbersome sample preparation steps and longer testing time. In addition, some coatings will produce a variety of corrosion products or phase change structures in complex environments, and it is difficult to quickly and accurately identify and track their changes using traditional methods.

[0004] Raman spectroscopy is gaining increasing attention in the field of materials research due to its fast testing speed, non-destructiveness, and insensitivity to moisture. However, when accurately identifying the damage evolution of multilayer or composite coatings, a single Raman detection method still has some problems: first, the Raman characteristic peaks of different coatings and corrosion products may overlap or be submerged in complex background noise; second, the lack of a systematic standard spectral database makes it difficult to quickly search and compare the product phases of new or composite coatings, and there is a lack of verification methods for test results; third, the existing data processing and analysis models are relatively limited, and it is difficult to take into account the accuracy of massive spectral data and the real-time analysis requirements.

[0005] Therefore, there is an urgent need for a coating damage evolution analysis method based on Raman spectroscopy. By combining multiple testing methods such as scanning electron microscopy, electrochemistry and Raman spectroscopy, a unified damage product database is established and advanced data preprocessing and model training methods are introduced to improve data reliability while improving the efficiency of coating damage monitoring and identification, and ultimately forming a set of systematic damage identification and performance evaluation methods. In view of the above difficulties and shortcomings, the present invention proposes a comprehensive analysis process and method, which can provide important technical support for damage identification and evaluation of coatings in actual service environments, and thus have a positive impact on the design and maintenance strategies of coatings. Summary of the invention

[0006] In view of the shortcomings of the prior art, an embodiment of the present invention aims to provide a coating damage evolution analysis method based on Raman spectroscopy to solve the problems in the above-mentioned background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The coating damage evolution analysis method based on Raman spectroscopy includes the following steps:

[0009] Step 1. Detect characteristic substances before coating damage: Use scanning electron microscopy (SEM) and energy dispersive spectrometer (EDS) to test the microscopic morphology and chemical composition of the coating, combine Raman spectroscopy technology to determine the characteristic peak position of the characteristic substances of the coating, and determine the characteristic substances before coating damage based on literature, databases, etc.

[0010] Step 2: Detect characteristic substances after coating damage: Use a knife to form clear "cross" scratches on the sample surface, use salt spray acceleration test to simulate extreme environment to prepare damaged coating samples, and repeat step 1 with SEM / EDS and Raman spectroscopy technology to identify the characteristic substances after damage.

[0011] Step 3: Electrochemical corrosion performance test before and after coating damage: Test the electrochemical corrosion performance during the coating damage evolution process to clarify the difference in electrochemical corrosion performance before and after coating damage.

[0012] Step 4. Establish a Raman spectral database of characteristic substances before and after damage: Purchase standard substances of characteristic substances that appear in the damage evolution process, use Raman spectrometer to collect 100 sets of spectral data for each standard substance, optimize spectral characteristic information through data preprocessing and expansion, increase the amount of Raman spectral data of each characteristic substance, and build a Raman spectral database of coating damage evolution.

[0013] Step 5: Training the coating damage evolution recognition model: Use the principal component analysis (PCA)-support vector machine (SVM) model and the random forest model to train the Raman spectroscopy database of coating damage evolution, and compare the application effectiveness of the models to obtain the classification and recognition model with the best performance.

[0014] Step 6. Design of coating damage evolution identification system: Combine the trained classification model with the best performance with the Raman spectrum preprocessing system to build an integrated coating damage evolution analysis system.

[0015] Step 7: Verification of system analysis results: Conduct electrochemical corrosion performance tests on the coating to verify the analysis results of the control coating damage evolution identification system and establish the correlation between "damage characterization-performance evolution".

[0016] As a further technical solution of the present invention, in step one, the qualitative characterization of the coating damage products can be combined with other structural or component characterization methods such as X-ray diffractometer and Fourier transform infrared spectroscopy.

[0017] As a further technical solution of the present invention, in step 2, before and after scratching, the surface cleanliness of the sample is strictly controlled, and a fixed load or constant depth scratching method is used to ensure the repeatability of the scratch depth / width, and the scratch morphology parameters are recorded. At the same time, the parameter conditions of the salt spray test are adjusted according to the coating application scenario, and other environmental simulations can also be used to damage the coating.

[0018] As a further technical solution of the present invention, in step three, the electrochemical impedance spectrum and the potentiodynamic polarization curve during the coating damage evolution process are tested by using an electrochemical workstation.

[0019] As a further technical solution of the present invention, in step 4, the Raman spectroscopy data expansion method is carried out by introducing a data enhancement algorithm, adding spectra with different noise levels, frequency shift perturbations, etc.

[0020] As a further technical solution of the present invention, in step five, both the PCA-SVM model and the random forest model use a grid search method to optimize hyperparameters, reduce fitting through leave-one-out cross-validation, and use confusion matrix, accuracy, sensitivity, specificity and other indicators to comprehensively compare model performance.

[0021] As a further technical solution of the present invention, in step 5, the use steps of the random forest model are as follows:

[0022] Step 1: Standardize the coating Raman spectrum data set to obtain Raman spectrum data in a unified format;

[0023] Step 2: using a bootstrap method or other multiple sub-sample sampling methods, randomly extract multiple sample subsets from the spectral data of the training set, and use them to train multiple decision trees respectively;

[0024] Step 3: Voting or weighted fusion of the classification results of each decision tree to obtain the final classification result of the random forest model and identify the coating components.

[0025] As a further technical solution of the present invention, in step six, the coating damage evolution analysis system includes coating Raman spectrum preprocessing and damage classification and identification.

[0026] As a further technical solution of the present invention, in step seven, the use steps of the electrochemical corrosion performance test are as follows:

[0027] Step 1: Using an electrochemical workstation, perform electrochemical impedance spectroscopy test and potentiodynamic polarization curve test on the samples before and after the coating is damaged;

[0028] Step 2: Record corrosion potential, current density, polarization resistance and other parameters;

[0029] Step 3: Compare and analyze the above parameters to evaluate the difference in corrosion resistance before and after coating damage, and cross-validate with Raman spectroscopy identification results.

[0030] Compared with the prior art, the present invention has the following beneficial effects:

[0031] (1) The present invention realizes rapid identification and accurate evaluation of the coating damage process. Compared with the traditional technical solution that relies on a single detection method, it can effectively reduce the detection blind area and improve the detection accuracy;

[0032] (2) The present invention incorporates the electrochemical corrosion performance test results into the coating damage evaluation system, which is mutually verified with the damage characteristic substances identified by Raman spectroscopy, and can more comprehensively and accurately reveal the damage evolution and failure mechanism of the coating under actual working conditions.

[0033] (3) The present invention provides a complete data preprocessing and intelligent analysis method, which overcomes the disadvantages of traditional Raman analysis that is susceptible to background disturbance and lacks systematic data processing, and significantly improves the recognition accuracy and reliability;

[0034] (4) The present invention constructs a Raman spectroscopy database applicable to a variety of environments and material systems by performing standardized testing and data expansion on typical substances that appear before and after coating damage, thereby achieving rapid identification of new coating materials or more complex corrosion products and having good scalability;

[0035] (5) The present invention integrates the optimized spectral preprocessing method with the trained optimal classification model to build an integrated coating damage evolution analysis system, which greatly improves the detection efficiency and application flexibility;

[0036] In order to more clearly illustrate the structural features and effects of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A flowchart of a coating damage evolution analysis method based on Raman spectroscopy provided in an embodiment of the present invention;

[0038] Figure 2 The Raman spectrum, morphology and element distribution of the coating surface provided by the embodiment of the present invention;

[0039] Figure 3The principal component analysis results and the confusion matrix diagram of the support vector machine model provided by the embodiment of the present invention;

[0040] Figure 4 A confusion matrix diagram of a random forest model provided by an embodiment of the present invention;

[0041] Figure 5 An operation interface of a Raman spectrum preprocessing system provided by an embodiment of the present invention;

[0042] Figure 6 An operation interface of a Raman spectrum classification and recognition system provided by an embodiment of the present invention; DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] The specific implementation of the present invention is described in detail below in conjunction with specific embodiments.

[0045] Example 1

[0046] like Figure 1 As shown, a coating damage evolution analysis method based on Raman spectroscopy provided as an embodiment of the present invention comprises the following steps:

[0047] Step 1. Detect characteristic substances before coating damage: Use scanning electron microscopy (SEM) and energy dispersive spectrometer (EDS) to test the microscopic morphology and chemical composition of the coating, combine Raman spectroscopy technology to determine the characteristic peak position of the characteristic substances of the coating, and determine the characteristic substances before coating damage based on literature, databases, etc.

[0048] Step 2: Detect characteristic substances after coating damage: Use a knife to form clear "cross" scratches on the sample surface, use salt spray acceleration test to simulate extreme environment to prepare damaged coating samples, and repeat step 1 with SEM / EDS and Raman spectroscopy technology to identify the characteristic substances after damage.

[0049] Step 3: Electrochemical corrosion performance test before and after coating damage: Test the electrochemical corrosion performance during the coating damage evolution process to clarify the difference in electrochemical corrosion performance before and after coating damage.

[0050] Step 4. Establish a Raman spectral database of characteristic substances before and after damage: Purchase standard substances of characteristic substances that appear in the damage evolution process, use Raman spectrometer to collect 100 sets of spectral data for each standard substance, optimize spectral characteristic information through data preprocessing and expansion, increase the amount of Raman spectral data of each characteristic substance, and build a Raman spectral database of coating damage evolution.

[0051] Step 5: Training the coating damage evolution recognition model: Use the principal component analysis (PCA)-support vector machine (SVM) model and the random forest model to train the Raman spectroscopy database of coating damage evolution, and compare the application effectiveness of the models to obtain the classification and recognition model with the best performance.

[0052] Step 6. Design of coating damage evolution identification system: Combine the trained classification model with the best performance with the Raman spectrum preprocessing system to build an integrated coating damage evolution analysis system.

[0053] Step 7: Verification of system analysis results: Conduct electrochemical corrosion performance tests on the coating to verify the analysis results of the control coating damage evolution identification system and establish the correlation between "damage characterization-performance evolution".

[0054] Figure 2 The results of coating inspection using SEM / EDS and Raman spectroscopy are shown:

[0055] The microstructure of the corrosion-resistant coating presents a uniformly distributed granular structure and fine pores. This structure may be formed by the composite of corrosion-resistant oxide particles (such as chromium oxide or aluminum oxide) and an organic coating matrix. The dense distribution between the particles can effectively block the penetration of corrosive media and enhance the barrier effect of the coating. In addition, the uniformity and fine pores of this microstructure help the coating to exhibit excellent corrosion resistance in marine or acidic environments;

[0056] EDS analysis of the coating shows that its elemental composition is complex and diverse, including strontium (31.42wt%), titanium (16.35wt%), chromium (15.83wt%), silicon (12.50wt%), magnesium (4.82wt%) and nickel (4.34wt%). Strontium compounds may exist in the form of strontium chromate, providing cathodic protection and effectively preventing the penetration of corrosive media. Titanium and chromium further enhance the anti-oxidation properties of the coating, while the presence of silicon and magnesium improves the weather resistance and mechanical stability of the matrix. The introduction of nickel enhances the durability of the coating in the marine environment;

[0057] from Figure 2 It can be seen that the Raman spectrum of the coating is at 867 cm -1 and 893cm -1There are obvious peaks at the bottom of the coating. Combining the literature and Raman spectrum database information, these peaks correspond to the characteristic peaks of strontium chromate (SrCrO4). SrCrO4 is a common anti-corrosion material, and its Raman activity characteristics are consistent with the test results, indicating that the coating has excellent corrosion resistance.

[0058] In step one, the characterization of coating damage products can also be combined with other structural or compositional characterization methods such as X-ray diffractometer and Fourier transform infrared spectroscopy.

[0059] In step two, before and after scratching, the surface cleanliness of the sample is strictly controlled, and a fixed load or constant depth scratch method is used to ensure the repeatability of the scratch depth / width, and the scratch morphology parameters are recorded. At the same time, the parameter conditions of the salt spray test are adjusted according to the coating application scenario, and other environmental simulations can also be used to damage the coating.

[0060] In step three, the electrochemical impedance spectroscopy and potentiodynamic polarization curves during the coating damage evolution process are tested by using an electrochemical workstation.

[0061] In step 4, Raman spectroscopy data expansion is carried out by introducing data enhancement algorithms, adding spectra with different noise levels, frequency shift perturbations, etc.

[0062] Example 2

[0063] As a preferred embodiment of the present invention, in step five, the PCA-SVM model and the random forest model are both optimized by grid search for hyperparameters, the leave-one-out cross-validation is used to reduce fitting, and the confusion matrix, accuracy, sensitivity, specificity, robustness and other indicators are used to comprehensively compare the model performance.

[0064] The PCA-SVM model is trained using the radial basis function kernel, and the optimal parameter combination is set to C = 10 and γ = 0.1 through grid search, which achieves accurate fitting of complex decision boundaries and can effectively distinguish multiple data categories. The ratio of the test set to the training set is 8:2;

[0065] Figure 3The results of using PCA to reduce the dimension of the Raman spectroscopy data set and the confusion matrix after importing the reduced-dimensional data into the SVM model training are shown. The results of principal component analysis show that the Raman spectral information accounts for 84.9%, of which the explained variance of principal component 1 is 66.33% and the explained variance of principal component 2 is 18.57%. It can be seen from the figure that different compounds show a relatively scattered distribution on the planes of principal component 1 and principal component 2, indicating that the Raman spectral characteristics of Cr2O3, Fe, Fe2O3, Mg2SiO4, SiO2, SrCrO4, TiO2 (rutile type) and TiO2 (anatase type) have significant differences, which are sufficient to show clear clustering partitions after dimensionality reduction. The values ​​of the rows and columns corresponding to Cr2O3 on the diagonal are much larger than those on the off-diagonal, indicating that SVM has a good recognition effect on Cr2O3 samples; for other categories, such as Fe2O3, Mg2SiO4, etc., it also has a high accuracy rate. Fe and Fe2O3 may have some characteristic peaks close to each other in the spectrum, which leads to misjudgment of very few samples, but the overall accuracy is still considerable; the test set results show that the PCA-SVM model achieved 100% accuracy, recall rate and F1 score in the classification task of 8 substances.

[0066] Since the random forest model and the PCA-SVM model are the same in terms of accuracy, sensitivity, and specificity, the more suitable model is selected by comparing the robustness of the two models. The coating Raman spectroscopy data was expanded by adding noise to 100 sets of Raman spectroscopy data, and imported into the above two models respectively. The results showed that the accuracy of the random forest model was 97%, while the accuracy of the PCA-SVM model was 99%, so the PCA-SVM model was selected for the subsequent design of the coating damage evolution classification and recognition system.

[0067] Example 3

[0068] As a preferred embodiment of the present invention, in step 5, the steps of using the random forest model are as follows:

[0069] Step 1: Standardize the coating Raman spectrum data set to obtain Raman spectrum data in a unified format;

[0070] Step 2: using a bootstrap method or other multiple sub-sample sampling methods, randomly extract multiple sample subsets from the spectral data of the training set, and use them to train multiple decision trees respectively;

[0071] Step 3: Voting or weighted fusion of the classification results of each decision tree to obtain the final classification result of the random forest model and identify the coating components.

[0072] Figure 4The confusion matrix of the random forest model is shown. As can be seen from the figure, all classified samples are correctly predicted, the diagonal elements are the maximum value, and the off-diagonal elements are 0, indicating that the RF model does not have any confusion between categories. Compared with other models, the RF model can more effectively deal with the nonlinear characteristics and potential noise of Raman spectroscopy data due to its ensemble learning characteristics and the diversity based on decision trees. The random forest model shows extremely high accuracy and stability in the classification of Raman spectroscopy data, and its test set classification results also reach 100% precision, recall rate and F1 score. This result shows the excellent performance of the RF model in multi-category classification tasks, and can effectively capture the characteristics of different substances in Raman spectroscopy.

[0073] For 8 types of standard substances (Cr2O3, Fe, Fe3O4, MgSiO4, SiO2, SrCrO4, TiO2 (rutile), TiO2 (anatase)), the classification index is 100%. This shows that there are no false positives in the classification prediction of all substance categories, and the model successfully identifies all test samples without missing any classifications, achieving a perfect balance between precision and recall in the classification task, demonstrating the reliability of the model.

[0074] Example 4

[0075] As a preferred embodiment of the present invention, in step six, the coating damage evolution analysis system includes coating Raman spectrum preprocessing and damage classification and identification. Before performing coating damage classification and identification, the Raman spectrum data is preprocessed, and then the processed Raman spectrum data is imported into the damage classification and identification system, which automatically performs PCA dimensionality reduction and standardization processing on the Raman spectrum data, and then uses the PCA-SVM model with the best comprehensive performance for classification and identification.

[0076] Figure 5 The Raman spectrum preprocessing system operation interface provided by the embodiment of the present invention. The system includes multiple Raman spectrum preprocessing functions such as spectral range selection, cosmic ray elimination, baseline calibration and smoothing. The user enters the minimum and maximum values ​​of the Raman shift according to personal needs, and clicks the "Update (range)" button after entering, and the Raman spectrum data graph within the selected range can be viewed in the right view area; according to the maximum value of the cosmic rays presented in the original data of the Raman spectrum, the interference of cosmic rays can be eliminated by selecting a suitable "threshold value"; the specific value that the user wants to set in the "polynomial order" part is entered, and the appropriate value is selected in the "regular coefficient" part below, and the baseline calibration of the Raman spectrum can be realized by clicking "Update (baseline)" on the right; the user can enter a specific value that is reasonably set in the "window length" part on the right side of the operation interface, and similarly select a specific value in the "order" part below, and click "Update (SG)" on the right to realize the smoothing of the Raman spectrum.

[0077] Figure 6 The operation interface of the Raman spectrum classification and identification system provided by the embodiment of the present invention. The user only needs to click the "Select File" button in the middle of the software interface and select the Raman spectrum file (xlsx) to be classified to achieve classification. The classification is demonstrated in the figure, and the "classification results are as follows: Iron oxide (Fe2O3) - component of the damaged absorbing coating; Tip: The results are for reference only, please combine experimental data for further analysis."

[0078] Example 5

[0079] As a preferred embodiment of the present invention, in step seven, the electrochemical corrosion performance test is performed as follows:

[0080] Step 1: Using an electrochemical workstation, perform electrochemical impedance spectroscopy test and potentiodynamic polarization curve test on the samples before and after the coating is damaged;

[0081] Step 2: Record corrosion potential, current density, polarization resistance and other parameters;

[0082] Step 3: Compare and analyze the above parameters to evaluate the difference in corrosion resistance before and after coating damage, and cross-validate with Raman spectroscopy identification results.

[0083] Compare the Raman spectrum classification and identification results of coating damage with the electrochemical corrosion performance results of coating damage. If the Raman spectrum classification and identification results of coating damage show that new substances that are only generated after damage appear in the coating, and the electrochemical corrosion performance of coating damage decreases, it means that the coating has been damaged and needs to be repaired in time.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A coating damage evolution analysis method based on Raman spectroscopy, characterized in that: The steps include: Step 1. Detect characteristic substances before coating damage: Use scanning electron microscopy (SEM) and energy dispersive spectrometer (EDS) to test the microscopic morphology and chemical composition of the coating, combine Raman spectroscopy technology to determine the characteristic peak position of the characteristic substances of the coating, and determine the characteristic substances before coating damage based on literature, databases, etc. Step 2: Detect characteristic substances after coating damage: Use a knife to form clear "cross" scratches on the sample surface, use salt spray acceleration test to simulate extreme environment to prepare damaged coating samples, and repeat step 1 with SEM / EDS and Raman spectroscopy technology to identify characteristic substances after damage. Step 3: Electrochemical corrosion performance test before and after coating damage: Test the electrochemical corrosion performance during the coating damage evolution process to clarify the difference in electrochemical corrosion performance before and after coating damage. Step 4. Establish a Raman spectral database of characteristic substances before and after damage: Purchase standard substances of characteristic substances that appear in the damage evolution process, use Raman spectrometer to collect 100 sets of spectral data for each standard substance, optimize spectral characteristic information through data preprocessing and expansion, increase the amount of Raman spectral data of each characteristic substance, and build a Raman spectral database of coating damage evolution. Step 5: Training the coating damage evolution recognition model: Use the principal component analysis (PCA)-support vector machine (SVM) model and the random forest model to train the Raman spectroscopy database of coating damage evolution, and compare the application effectiveness of the models to obtain the classification and recognition model with the best performance. Step 6. Design of coating damage evolution identification system: Combine the trained classification model with the best performance with the Raman spectrum preprocessing system to build an integrated coating damage evolution analysis system. Step 7: Verification of system analysis results: Conduct electrochemical corrosion performance tests on the coating to verify the analysis results of the control coating damage evolution identification system and establish the correlation between "damage characterization-performance evolution".

2. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In the step 1, the qualitative characterization of the coating damage products can be combined with other structural or component characterization methods such as X-ray diffractometer and Fourier transform infrared spectroscopy.

3. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In the step 2, before and after scratching, the surface cleanliness of the sample is strictly controlled, and a fixed load or constant depth scratching method is used to ensure the repeatability of the scratch depth / width, and the scratch morphology parameters are recorded. At the same time, the parameter conditions of the salt spray test are adjusted according to the coating application scenario, and other environmental simulations can also be used to damage the coating.

4. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In the step three, an electrochemical workstation is used to test the electrochemical impedance spectrum and potentiodynamic polarization curve during the coating damage evolution process.

5. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In the step 4, the Raman spectroscopy data expansion is carried out by introducing a data enhancement algorithm, adding spectra with different noise levels, frequency shift perturbations, etc.

6. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In step 5, both the PCA-SVM model and the random forest model use a grid search method to optimize hyperparameters, cross-validation is used to reduce fitting, and confusion matrix, accuracy, sensitivity, specificity and other indicators are used to comprehensively compare model performance.

7. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In step 5, the steps for using the random forest model are as follows: Step 1: Standardize the coating Raman spectrum data set to obtain Raman spectrum data in a unified format; Step 2: using a bootstrap method or other multiple sub-sample sampling methods, randomly extract multiple sample subsets from the spectral data of the training set, and use them to train multiple decision trees respectively; Step 3: Voting or weighted fusion of the classification results of each decision tree to obtain the final classification result of the random forest model and identify the coating components.

8. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In step six, the coating damage evolution analysis system includes coating Raman spectrum preprocessing and damage classification and identification.

9. The coating damage evolution analysis method based on Raman spectroscopy according to claim 1, characterized in that: In step 7, the steps for using the electrochemical corrosion performance test are as follows: Step 1: Using an electrochemical workstation, perform electrochemical impedance spectroscopy test and potentiodynamic polarization curve test on the samples before and after the coating is damaged; Step 2: Record corrosion potential, current density, polarization resistance and other parameters; Step 3: Compare and analyze the above parameters to evaluate the difference in corrosion resistance before and after coating damage, and cross-validate with Raman spectroscopy identification results.

Images