Aluminum alloy atmospheric corrosion prediction method and system based on machine learning

Abstract

The invention provides an aluminum alloy atmospheric corrosion prediction method and system based on machine learning, and relates to the technical field of aluminum alloy corrosion prediction. Comprising the following steps: collecting aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy component characteristics and aluminum alloy corrosion rate to obtain an aluminum alloy atmospheric corrosion initial data set; performing screening and dimension reduction on the data in the initial data set to obtain an aluminum alloy atmospheric corrosion prediction data set; based on the aluminum alloy atmospheric corrosion prediction data set, the built CNN-SVM model is trained; and acquiring environment data of the to-be-tested aluminum alloy, and predicting the atmospheric corrosion rate of the to-be-tested aluminum alloy based on the trained CNN-SVM model. According to the method, accurate feature screening and short-time-consumption model training are carried out by utilizing limited experimental data, the atmospheric corrosion of the aluminum alloy is predicted at high precision, the cost input of a traditional experiment is reduced, the time cost is saved, and the prediction precision and efficiency are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of aluminum alloy corrosion prediction, and particularly relates to an aluminum alloy atmospheric corrosion prediction method and system based on machine learning. BACKGROUND

[0002] Aluminum alloy has the characteristics of high specific strength, high specific stiffness and excellent processing performance, and is widely used in the fields of aviation, aerospace, shipbuilding and electronics, and is usually used as a structural material during service. In the atmospheric environment, aluminum alloy is subjected to the combined action of corrosion media, and corrosion occurs. Atmospheric corrosion of aluminum alloy may cause failure of structural components, causing significant losses, and aluminum alloy corrosion evaluation and prediction are of great significance for the selection and service of aluminum alloy.

[0003] Due to the small amount of experimental cumulative data and high experimental cost, researchers try to predict the corrosion behavior of aluminum alloy in the atmospheric environment by simulation, data mining and other methods, but often face problems such as long time consumption, high calculation difficulty and low precision. SUMMARY

[0004] To overcome the above-mentioned deficiencies of the prior art, the present application provides an aluminum alloy atmospheric corrosion prediction method and system based on machine learning, which designs an advanced machine learning algorithm to study the relationship between the corrosion of aluminum alloy in the atmospheric environment and the environmental impact factors and the composition of aluminum alloy, and uses feature screening to screen out key influence features affecting the atmospheric corrosion of aluminum alloy. Based on this, a machine learning model with high precision and high generalization ability is established, which can accurately predict the atmospheric corrosion rate of aluminum alloy.

[0005] To achieve the above-mentioned purpose, one or more embodiments of the present application provide the following technical solutions: The first aspect of the present application provides an aluminum alloy atmospheric corrosion prediction method based on machine learning.

[0006] The aluminum alloy atmospheric corrosion prediction method based on machine learning comprises the following steps: Collecting aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy composition characteristics and aluminum alloy corrosion rate to obtain an initial aluminum alloy atmospheric corrosion data set; Filtering and reducing the data in the initial data set to obtain an aluminum alloy atmospheric corrosion prediction data set; Training the built CNN-SVM model based on the aluminum alloy atmospheric corrosion prediction data set; Obtaining the environmental data of the aluminum alloy to be tested, and predicting the atmospheric corrosion rate of the aluminum alloy to be tested based on the trained CNN-SVM model.

[0007] The second aspect of the present application provides an aluminum alloy atmospheric corrosion prediction system based on machine learning.

[0008] The aluminum alloy atmospheric corrosion prediction system based on machine learning comprises: A data acquisition module is configured to collect aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy component characteristics and aluminum alloy corrosion rates, to obtain an initial aluminum alloy atmospheric corrosion data set; A data dimension reduction module is configured to filter and reduce the data in the initial data set, to obtain an aluminum alloy atmospheric corrosion prediction data set; A model training module is configured to train a CNN-SVM model based on the aluminum alloy atmospheric corrosion prediction data set; A prediction module is configured to obtain environmental data in which the aluminum alloy to be tested is located, and to predict the atmospheric corrosion rate of the aluminum alloy to be tested based on the trained CNN-SVM model. The third aspect of the present application provides a computer readable storage medium having a program stored thereon, wherein the program is executed by a processor to implement the steps of the aluminum alloy atmospheric corrosion prediction method based on machine learning according to the first aspect of the present application.

[0009] The fourth aspect of the present application provides an electronic device comprising a memory, a processor and a program stored on the memory and executable on the processor, wherein the processor executes the program to implement the steps of the aluminum alloy atmospheric corrosion prediction method based on machine learning according to the first aspect of the present application.

[0010] The above one or more technical solutions have the following beneficial effects: The present application provides an aluminum alloy atmospheric corrosion prediction method and system based on machine learning, which uses machine learning models to predict the corrosion of aluminum alloys in atmospheric environments based on aluminum alloy components and atmospheric environmental factors. The method can use limited experimental data to accurately select features and train models with short time consumption, accurately predict aluminum alloy atmospheric corrosion, reduce the cost investment of traditional experiments, save time cost, and provide prediction accuracy and efficiency.

[0011] The prediction method of the present application uses advanced machine learning methods, uses limited data resources, and can efficiently predict the atmospheric corrosion rate of aluminum alloys. It has the advantages of high precision and strong generalization, and can be used for prediction in different environments and conditions. It is of great significance to optimize the performance evaluation of aluminum alloys and prolong the service life of aluminum alloys.

[0012] The advantages of the additional aspects of the present application will be partially given in the following description, partially will become apparent from the following description, or will be understood by the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings, which form a part of this specification, are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification. The embodiments of the application, and their

[0014] Figure 1 Data preprocessing and correlation calculation flowchart of the application.

[0015] Figure 2 Feature screening based on Pearson correlation coefficient value of the application.

[0016] Figure 3 Scatter plot of the predicted value and the experimental value of different machine learning models.

[0017] Figure 4 Scatter plot of the predicted value and the experimental value comparison of the hybrid model CNN-SVM.

[0018] Figure 5 Performance test chart of CNN-SVM.

[0019] Figure 6 Method flowchart of the first embodiment of the application. DETAILED DESCRIPTION

[0020] It should be noted that the following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0021] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the application.

[0022] The embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0023] Embodiment one This embodiment aims to overcome the shortcomings of traditional experiments, use advanced computing methods, and provide a new, high-computing-efficiency, high-precision aluminum alloy atmospheric corrosion prediction method, which can reduce the loss and prolong the service life of aluminum alloy, and has important significance for the field of material science and engineering applications.

[0024] As shown in Figure 1 The aluminum alloy atmospheric corrosion prediction method based on machine learning disclosed in this embodiment comprises the following steps: Collecting aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy composition characteristics and aluminum alloy corrosion rate to obtain an initial aluminum alloy atmospheric corrosion data set; The data in the initial dataset were filtered and dimensionality reduced to obtain the aluminum alloy atmospheric corrosion prediction dataset. The CNN-SVM model was trained based on the aluminum alloy atmospheric corrosion prediction dataset. The environmental data of the aluminum alloy under test is obtained, and the atmospheric corrosion rate of the aluminum alloy under test is predicted based on the trained CNN-SVM model.

[0025] The technical solution of this embodiment will be explained in detail below.

[0026] The embodiments of the present invention are implemented through the following technical solutions, including the following steps: Step S1: Conduct a one-year exposure test to collect atmospheric corrosion data of aluminum alloys and obtain an initial dataset of atmospheric corrosion of aluminum alloys.

[0027] The atmospheric corrosion data for the aluminum alloy includes environmental meteorological data, aluminum alloy composition characteristics, and aluminum alloy corrosion rate. The mass fractions of Fe and Si in the aluminum alloy sample are used as the aluminum alloy composition characteristics.

[0028] The data obtained specifically includes: first, the corrosion rate of aluminum alloys obtained based on weight loss analysis; and second, environmental meteorological data, including average temperature, relative humidity, chloride ion deposition rate, SO2 concentration, NO2 concentration, and PM2.5.

[0029] Furthermore, step S1 specifically includes the following steps: Step S11: Literature review, collecting aluminum alloy composition data, atmospheric environment data, and aluminum alloy atmospheric corrosion data; after the aluminum alloy undergoes a one-year exposure test and is derusted, the corrosion rate of the exposed sample is calculated using the weight loss method. Literature search sources include the databases Sciencedirect, Springer, CKNI, and Web of Science.

[0030] Step S12: Data preprocessing: such as Figure 1 As shown, outlier detection, standardization, and missing value imputation are performed on the data in sequence to improve data quality and consistency. The outlier handling involves eliminating outliers caused by experimental or recording errors based on the literature review results in step S11. The standardization process aims to eliminate the influence of different units of measurement by normalizing the data to the range of 0 to 1, and calculating the minimum value X of the original data for each feature. min and maximum value X max And convert each data point to X i-new = , where X iis the original value before normalization processing. The missing value filling adopts spline interpolation, which interpolates between discrete data points by a low-order polynomial.

[0031] Step S13: data set construction: the preprocessed data needs to be effectively integrated to establish a data set for machine learning training, and the data set construction specifically includes: 1) data integration: aligning the data of different dimensions and units according to the characteristics, and integrating them in a unified format; 2) data division: randomly extracting the data set into a training set and a test set, the training set is used to train the model, and the test set is used to test the model performance.

[0032] Step S2: screening and dimension reduction of features according to gray correlation coefficient and Pearson correlation coefficient, to obtain an aluminum alloy atmospheric corrosion prediction data set.

[0033] Specifically includes: The average temperature, relative humidity, chloride ion deposition rate, SO2 concentration, NO2 concentration, PM2.5, mass fraction of Fe and mass fraction of Si are respectively taken as the dimensions of the initial data set; The gray correlation coefficient of each dimension data and the aluminum alloy atmospheric corrosion rate is calculated; The Pearson correlation coefficient between all dimension data is calculated; From all the dimensions, the dimensions whose gray correlation coefficient and Pearson correlation coefficient both meet the set threshold are selected as the final dimensions, to form the aluminum alloy atmospheric corrosion prediction data set.

[0034] More specifically, the step S2 includes: Step S21: gray correlation coefficient calculation, calculating the gray correlation coefficient of each feature and the aluminum alloy atmospheric corrosion rate.

[0035] The gray correlation calculation is shown in the following formulas (2)-(4). Taking the aluminum alloy corrosion rate as the mother sequence and the average temperature, relative humidity, salt particle deposition amount, NO2 content, SO2 content, Fe component mass fraction and Si component mass fraction as the sub-sequences, the gray correlation degrees of each feature to the aluminum alloy corrosion rate are calculated respectively. According to the calculation, the gray correlation degrees are respectively: 0.5961, 0.5465, 0.7512, 0.7986, 0.7959, 0.5877, 0.7438, all greater than 0.5, which can indicate that the seven features have a greater impact on the aluminum alloy corrosion rate.

[0036] (2) (3) (4) In the formula, - Grey correlation degree; - Discrimination coefficient equal to 0.5; - Mother sequence data; - Sub-sequence.

[0037] Step S22: Pearson correlation coefficient calculation, calculate the correlation coefficient between all features.

[0038] The formula for calculating the Pearson correlation coefficient is formula (5).

[0039] (5) In the formula, - Pearson correlation coefficient; , - Two types of environmental factor variables; , - Mean value of the variable.

[0040] Figure 2 The Pearson correlation coefficient between the environmental features is shown in the figure, which analyzes the correlation coefficient between temperature, relative humidity, chloride deposition rate, PM2.5, NO2 content and SO2 content features pairwise. The closer the correlation coefficient is to 1, the stronger the correlation. PM2.5 and NO2 content and SO2 content have very strong correlation.

[0041] Step S23: Feature selection: According to the analysis results of the grey correlation coefficient, the features that have the most significant impact on the prediction of aluminum alloy atmospheric corrosion are selected. The importance of the feature is evaluated by the absolute size of its grey correlation coefficient. A larger grey correlation coefficient means that the feature is more important in the model. In order to avoid the interaction of highly correlated features in the model, causing overfitting, according to the Pearson correlation coefficient, only 1 to 2 features with larger Pearson correlation coefficients are selected from the selected features.

[0042] The environmental features mainly include average temperature, relative humidity, PM2.5, salt particle deposition, NO2 content and SO2 content. In addition, there are two component features: mass fraction of Fe and Si.

[0043] The Pearson correlation coefficient between the environmental features is shown in Figure 2 , where PM2.5 and NO2 content and SO2 content have very strong correlation. Through this step, the data of PM2.5 is filtered out. Therefore, the dataset composed of the five environmental feature variables: average temperature, relative humidity, salt particle deposition, NO2 content, SO2 content and the two component features: mass fraction of Fe and Si is constructed as input, and the aluminum alloy corrosion rate is constructed as output.

[0044] Step S3: According to the prediction data set, the CNN-SVM model is built and trained, and the results are compared with single machine learning algorithms widely used in data prediction to obtain the preliminary aluminum alloy atmospheric corrosion prediction model.

[0045] The construction of the CNN-SVM model mainly includes two parts, namely model training and testing. The data set is divided into two independent training and testing sets. The implementation process mainly includes data preprocessing, CNN network building, CNN optimizer and loss function selection, and feature extraction from the CNN network for SVM model creation and training.

[0046] CNN mainly consists of input layer, convolution layer, pooling layer and fully connected layer, etc. In the proposed architecture, three one-dimensional convolution layers are adopted. Due to the small size of the data, only one pooling layer is adopted; for small-scale data sets, overfitting phenomenon is easy to occur, and operations such as regularization are needed, so batch normalization layer (BN) and dropout layer are adopted; and Relu activation layer is adopted.

[0047] Through steps S1 and S2, the data set is obtained, and then the average temperature, relative humidity, salt particle deposition amount, NO2 content, SO2 content, Fe component mass fraction and Si component mass fraction are input into the built CNN-SVM model. The convolution kernel size is 3*1, and for small-scale data sets, BN regularization processing is performed; ReLU activation function is used: "f(x)=max(0,x)"; due to the small size of the data, only one pooling layer is adopted; Dropout is used to randomly ignore a part of neurons to prevent overfitting of the neural network and fall into local optimal problem. Finally, the fully connected layer is used to map the features to the sample label space.

[0048] The SVM type is e-SVR, and the value of the loss function p in e-SVR is 0.01. The kernel function is RBF function. SVM adopts 5-fold cross-validation to obtain the optimal parameters of g parameter (gamma function setting in kernel function) and c parameter (penalty coefficient).

[0049] Further optimization of the hyperparameters of CNN, selection of optimizers (SGD, SGDM, ADAM, RMSProp), dropout regularization (0.1, 0.2, 0.4, 0.8), maximum number of training times (800, 1000, 1500, 2000), initial learning rate (0.05, 0.01, 0.001). Under the condition of keeping other hyperparameters unchanged, the training set and test set are randomly allocated from the data set, and finally the optimal hyperparameter combination is obtained.

[0050] After optimization, all hyperparameters are as follows: optimizer: SGDM; dropout regularization: 0.1 (keep 90% of connections); maximum number of training times: 1500; initial learning rate: 0.01.

[0051] In the model, 80% of the data is used for training, and after the training is completed, the remaining 20% of the untrained data is used as input. The feature parameters include average temperature, relative humidity, salt particle deposition amount, NO2 content, SO2 content, Fe component mass fraction, and Si component mass fraction. After the CNN-SVM model, the predicted aluminum alloy corrosion rate is obtained. The predicted value obtained by the model is compared with the actual value, and MRE, RMSE and R2 are used to evaluate the model precision.

[0052] A single machine learning algorithm widely used in prediction tasks is selected, including KNN, BP neural network, SVM, gradient boosting regression and the hybrid algorithm CNN-SVM provided in the embodiment, and a comparative experiment is performed. Figure 3 is a scatter plot of the test set prediction results and experimental values of different single machine learning models in the embodiment. Figure 3 In the embodiment, the data set composed of five environmental characteristic variables of average temperature, relative humidity, salt particle deposition amount, NO2 content and SO2 content and two component characteristics of Fe and Si mass fraction is input, and the aluminum alloy corrosion rate is output; the data points in the figure are distributed around the straight line y=x, and the closer to the straight line, the higher the prediction accuracy of the model.

[0053] Figure 4 is a scatter plot of the test set prediction results and experimental values of the hybrid machine learning model CNN-SVM used in the embodiment, and the data points are densely distributed around y=x, indicating the high precision value of the model.

[0054] Figure 5 is the precision verification of the CNN-SVM model with significantly higher precision in the embodiment, which is used to show the high precision value and robustness of the model. The figure shows the test results of the model by repeatedly randomly allocating the data set, which shows the high precision and good robustness of the CNN-SVM model. The MRE is less than 0.15, which shows that the model has high precision in the prediction of aluminum alloy atmospheric corrosion, and can be widely used in different data sets and can be extended to the prediction of aluminum alloy atmospheric corrosion rate under different conditions.

[0055] Step S4: According to the prediction model used in the embodiment, the model generalization ability and robustness test are performed.

[0056] Further, step S4 specifically includes the following steps: Step S41: cross-validation, the generalization ability of different models is verified by five-fold cross-validation technology; Step S42: performance evaluation, the results of each cross-validation are summarized to determine the coefficient R 2 , the value of the root mean square error MRE to evaluate the performance of the model; Step S43: based on the performance evaluation results, it is explained that the model structure designed in this embodiment has the highest R 2 value and the lowest MRE value.

[0057] Embodiment two The embodiment discloses an aluminum alloy atmospheric corrosion prediction system based on machine learning.

[0058] The aluminum alloy atmospheric corrosion prediction system based on machine learning comprises: A data acquisition module configured to collect aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy component characteristics and aluminum alloy corrosion rate, to obtain an initial aluminum alloy atmospheric corrosion data set; A data dimension reduction module configured to filter and reduce the data in the initial data set to obtain an aluminum alloy atmospheric corrosion prediction data set; A model training module configured to train a CNN-SVM model built based on the aluminum alloy atmospheric corrosion prediction data set; A prediction module configured to obtain environmental data where the aluminum alloy to be tested is located, and predict the atmospheric corrosion rate of the aluminum alloy to be tested based on the trained CNN-SVM model.

[0059] Embodiment three The purpose of this embodiment is to provide a computer readable storage medium.

[0060] The computer readable storage medium has a computer program stored thereon, and the program is executed by a processor to realize the steps in the aluminum alloy atmospheric corrosion prediction method based on machine learning as described in Embodiment 1 of the present disclosure.

[0061] Embodiment four The purpose of this embodiment is to provide an electronic device.

[0062] The electronic device comprises a memory, a processor, and a program stored on the memory and executable on the processor, and the processor executes the program to realize the steps in the aluminum alloy atmospheric corrosion prediction method based on machine learning as described in Embodiment 1 of the present disclosure.

[0063] The steps involved in the apparatuses of the above embodiments two, three and four correspond to the method of embodiment one, and the specific implementation can refer to the relevant description of embodiment one. The term "computer readable storage medium" should be understood as including a single medium or multiple media of one or more instruction sets; it should also be understood as including any medium capable of storing, encoding or carrying the instruction set for execution by the processor and causing the processor to perform any of the methods in the present application.

[0064] Those skilled in the art should understand that each module or step of the present application described above can be realized by a general computer device, alternatively, they can be realized by program codes executable by a computing device, so that they can be stored in a storage device for execution by a computing device, or they can be respectively made into each integrated circuit module, or a plurality of modules or steps among them can be made into a single integrated circuit module to realize. The present application is not limited to any specific combination of hardware and software.

[0065] Although the specific embodiments of the present application are described above in combination with the drawings, it is not a limitation on the protection scope of the present application, and those skilled in the art should understand that various modifications or changes made by those skilled in the art on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. A method for predicting atmospheric corrosion of aluminum alloys based on machine learning, characterized by, The method comprises the following steps: Collecting aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy component characteristics and aluminum alloy corrosion rate to obtain an initial aluminum alloy atmospheric corrosion data set; Filtering and reducing the data in the initial data set to obtain an aluminum alloy atmospheric corrosion prediction data set; Training a CNN-SVM model based on the aluminum alloy atmospheric corrosion prediction data set; Obtaining environmental data of the aluminum alloy to be tested and predicting the atmospheric corrosion rate of the aluminum alloy to be tested based on the trained CNN-SVM model.

2. The machine learning based aluminum alloy atmospheric corrosion prediction method of claim 1, wherein, The aluminum alloy atmospheric corrosion data includes environmental meteorological data, aluminum alloy component characteristics and aluminum alloy corrosion rate, and specifically includes: Performing a specified time exposure test on the aluminum alloy sample, and calculating the corrosion rate of the sample by the weight loss method after rust removal; Obtaining environmental meteorological data in the exposure test of the aluminum alloy sample, wherein the environmental meteorological data specifically includes average temperature, relative humidity, chloride ion deposition rate, SO2 concentration, NO2 concentration and PM2.5; Taking the mass fraction of Fe and Si in the aluminum alloy sample as the aluminum alloy component characteristics; Performing data preprocessing on the aluminum alloy atmospheric corrosion data, including outlier detection, standardization processing and missing value filling.

3. The machine learning based aluminum alloy atmospheric corrosion prediction method of claim 2, wherein, Filtering and reducing the data in the initial data set by using the grey correlation coefficient and the Pearson correlation coefficient, specifically including: Taking the average temperature, relative humidity, chloride ion deposition rate, SO2 concentration, NO2 concentration, PM2.5, mass fraction of Fe and mass fraction of Si as the dimensions of the initial data set, respectively; Calculating the grey correlation coefficient of each dimension data and the aluminum alloy atmospheric corrosion rate; Calculating the Pearson correlation coefficient between all dimension data; From all the dimensions, filtering out the dimensions whose grey correlation coefficient and Pearson correlation coefficient both meet the set threshold value as the final dimensions to form the aluminum alloy atmospheric corrosion prediction data set.

4. The machine learning based aluminum alloy atmospheric corrosion prediction method of claim 3, wherein, The final dimensions include average temperature, relative humidity, chloride ion deposition rate, SO2 concentration, NO2 concentration, mass fraction of Fe and mass fraction of Si.

5. The machine learning based aluminum alloy atmospheric corrosion prediction method of claim 1, wherein, Training the CNN-SVM model based on the aluminum alloy atmospheric corrosion prediction data set, specifically including: Inputting the data in the aluminum alloy atmospheric corrosion prediction data set into the CNN, sequentially passing through the input layer, convolution layer, pooling layer, normalization layer, dropout layer and full connection layer for feature extraction, and inputting the extracted features into the SVM for training, taking the predicted aluminum alloy corrosion rate as the output, to obtain the trained CNN-SVM model.

6. The machine learning based aluminum alloy atmospheric corrosion prediction method of claim 1, wherein, Obtaining the environmental data of the aluminum alloy to be tested and predicting the atmospheric corrosion rate of the aluminum alloy to be tested based on the trained CNN-SVM model, specifically including: Obtaining the average temperature, relative humidity, chloride ion deposition rate, SO2 concentration and NO2 concentration in the environment where the aluminum alloy to be tested is located as the environmental data of the aluminum alloy to be tested; Inputting the environmental data of the aluminum alloy to be tested and the mass fraction of Fe and Si into the CNN-SVM model together to predict the atmospheric corrosion rate of the aluminum alloy to be tested.

7. The machine learning based aluminum alloy atmospheric corrosion prediction method of claim 3, wherein, In calculating the grey correlation coefficient, dimensions with the grey correlation coefficient greater than a first preset threshold are retained; in calculating the Pearson correlation coefficient, dimensions with the Pearson correlation coefficient greater than a second preset threshold are screened out.

8. A machine learning based aluminum alloy atmospheric corrosion prediction system characterized by, Comprise: The data acquisition module is configured to collect aluminum alloy atmospheric corrosion data including environmental meteorological data, aluminum alloy component characteristics and aluminum alloy corrosion rates to obtain an aluminum alloy atmospheric corrosion initial data set; The data dimension reduction module is configured to filter and reduce the data in the initial data set to obtain an aluminum alloy atmospheric corrosion prediction data set; The model training module is configured to train the built CNN-SVM model based on the aluminum alloy atmospheric corrosion prediction data set; The prediction module is configured to obtain environmental data of the aluminum alloy to be tested and predict the atmospheric corrosion rate of the aluminum alloy to be tested based on the trained CNN-SVM model.

9. A computer-readable storage medium having stored thereon a program, characterized in that, The program is executed by the processor to implement the steps in the aluminum alloy atmospheric corrosion prediction method based on machine learning in any one of claims 1-7.

10. An electronic device comprising a memory, a processor, and a program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps in the aluminum alloy atmospheric corrosion prediction method based on machine learning in any one of claims 1-7.

Images