Corrosion environment factor and liquid film thickness correlation model construction method, prediction method and system

Abstract

The invention relates to a corrosion environment factor and liquid film thickness correlation model construction method, a prediction method and a system thereof, and belongs to the technical field of metal material corrosion prediction. The method comprises the following steps: obtaining a liquid film thickness based on corrosion environment factor data of a target area; constructing a data set which takes the corrosion environment factor data as an input feature and takes the liquid film thickness as an output feature; carrying out training and precision verification on the neural network based on the data set; and taking the neural network after training and precision verification as a corrosion environment factor and liquid film thickness correlation model. According to the method, the metal surface liquid film thickness is used as an intermediate key parameter and is used as a core driving factor of a corrosion occurrence and development mechanism, and the correlation model is constructed around the liquid film thickness, so that the real-time liquid film thickness is obtained, namely, the dynamic prediction of the liquid film thickness is realized.

Description

Technical Field

[0001] This invention belongs to the field of metal material corrosion prediction technology, and specifically relates to the construction method, prediction method and system of the correlation model between corrosion environment factors and liquid film thickness. Background Technology

[0002] With marine equipment, aircraft, and coastal infrastructure operating in humid, high-salt environments for extended periods, their critical metal components are exposed to complex atmospheric corrosion environments, making them susceptible to performance degradation due to corrosive media. The formation and thickness of the liquid film are key factors determining the occurrence and rate of atmospheric corrosion, influenced by a combination of environmental factors such as temperature, humidity, chloride ion concentration, and aerosol migration. However, existing methods for predicting liquid film thickness are mostly experimental, making them ill-suited to dynamic and complex environmental changes, especially in highly corrosive areas such as tropical islands and ports, where effective time-series prediction methods are lacking. Furthermore, traditional methods struggle to integrate multiple meteorological and corrosive factors and have limited ability to describe the complex nonlinear relationships between them.

[0003] For example, CN115265387A discloses a dynamic liquid film thickness measurement system based on fluorescence intensity, which includes a liquid film generation module, an optical illumination module, an image acquisition module, and an image processing module. This method, based on fluorescence intensity, offers a simple and low-cost liquid film thickness measurement system suitable for laboratory applications. However, its technology relies on specific optical devices, imaging conditions, and the fluorescence response characteristics of the liquid film, making it difficult to adapt to real-time and continuous measurement of liquid films in complex natural environments. Furthermore, it cannot establish a direct correlation with environmental factors and lacks modeling and predictive capabilities. Another example is the device and method for simultaneously measuring liquid film concentration and thickness disclosed in CN111948082A. This system mainly consists of two laser light sources with different wavelengths, a wavelength division multiplexer, a collimator, a liquid film carrier, a wavelength division multiplexer, two spectral acquisition components, and a computer. This dual-wavelength laser method for simultaneously measuring liquid film concentration and thickness possesses high accuracy and multi-parameter measurement capabilities. However, the device is highly dependent on the calibration and stability of the optical system, limiting its application to controlled experimental conditions. It cannot be continuously deployed in open environments such as oceans and cities, lacks data-driven prediction capabilities, and does not address the dynamic evolution mechanism between the liquid film and environmental factors. Another example is the material corrosion simulation modeling method based on real-time data acquisition disclosed in CN118228510 A. This method directly constructs the statistical relationship between environmental indicators and corrosion rates based on real-time data acquisition, possessing strong engineering application value. However, its core modeling method still relies primarily on traditional statistical methods such as linear regression and grayscale models, limiting its ability to model the nonlinear coupling characteristics between multiple variables. Furthermore, it lacks modeling of the intermediate physical mechanisms of corrosion formation processes (such as liquid film formation), reducing the model's physical interpretability and generalization ability. Yet another example is the atmospheric corrosion model construction method and related device for power grid equipment materials disclosed in CN118228510A. This atmospheric corrosion modeling method for power grid equipment focuses on power grid equipment and uses sensor data to establish an environmental dose response function, which has strong industry relevance. However, this method still relies mainly on function fitting and lacks deep time-series modeling capabilities, making it difficult to capture the historical dependence characteristics of the corrosion process. Furthermore, the model is highly dependent on existing measured corrosion rate values ​​for training and does not model precursor processes such as liquid film formation, limiting the model's generalizability and adaptability.

[0004] In summary, existing technologies for modeling liquid film thickness lack time-series predictive capabilities and struggle to handle dynamic changes in environmental factors. Most technologies rely on experimental conditions, making them difficult to apply to natural service environments. There is a lack of physical mechanism-driven intermediate variable modeling; while direct modeling of corrosion rates is convenient, its physical interpretability and adaptability are poor. Traditional statistical modeling methods have significant limitations in handling nonlinear and time-series characteristics. Therefore, a dynamic corrosion prediction scheme that can integrate multi-source environmental data, possesses time-series modeling capabilities, and uses liquid film thickness as the core variable remains lacking. Summary of the Invention

[0005] To address the above problems, this invention provides a method for constructing a correlation model between corrosion environmental factors and liquid film thickness, a prediction method, and a system thereof.

[0006] The first objective of this invention is to provide a method for constructing a correlation model between corrosion environment factors and liquid film thickness, comprising: The liquid film thickness is obtained based on corrosion environment factor data of the target area; Construct a dataset with corrosion environment factor data as input features and liquid film thickness as output feature; Based on the dataset, the neural network is trained and its accuracy is verified. The neural network, after training and accuracy verification, was used as a correlation model between corrosion environment factors and liquid film thickness.

[0007] In a specific embodiment of the present invention, the corrosive environmental factors include relative humidity, temperature, atmospheric pressure, chloride ion content, wind speed, wind direction, distance from the coast, altitude, ammonia concentration, sulfur dioxide concentration, and nitrogen dioxide concentration.

[0008] In a specific embodiment of the present invention, obtaining the liquid film thickness based on corrosion environment factor data of the target area includes: The salt spray deposition rate is calculated based on the chloride ion content in the corrosion environment factor data. The liquid film thickness was obtained based on the relative humidity and salt spray deposition rate from the corrosion environment factor data.

[0009] In a specific embodiment of the present invention, obtaining the liquid film thickness based on the relative humidity and salt spray deposition rate from the corrosion environment factor data includes: Determine whether the relative humidity of the environment is greater than the critical humidity of the metal surface; Based on the judgment that the relative humidity of the environment is greater than the critical humidity of the metal surface, the liquid film thickness is calculated according to the salt spray deposition rate and the relative humidity of the environment.

[0010] In a specific embodiment of the present invention, the formula for calculating the liquid film thickness is as follows:

[0011] in, FT For liquid film thickness, DD Salt spray deposition rate, RH Relative humidity, a 1. a 2. a 3. b 1. b 2. b 3 are all constants.

[0012] In a specific embodiment of the present invention, using corrosion environment factor data as input features includes: Correlation analysis of liquid film thickness was performed on corrosion environmental factor data; Based on the correlation analysis results of liquid film thickness, key corrosion environment factor data were selected; Key corrosion environmental factor data were used as input features.

[0013] In a specific embodiment of the present invention, the neural network is a long short-term memory neural network, which includes an input layer, two hidden layers and an output layer, and is used to fit the nonlinear temporal mapping relationship between input environmental factors and liquid film thickness.

[0014] In a specific embodiment of the present invention, the step of training and accuracy verification of the neural network based on the dataset includes: The neural network is trained based on the training sample set in the dataset; The trained neural network is validated using the test sample set in the dataset; Based on the verification results, calculate the accuracy index data; Based on the accuracy index data, determine whether the trained neural network effectively captures the dynamic relationship between liquid film thickness and environmental factors. Since the trained neural network failed to effectively capture the dynamic relationship between liquid film thickness and environmental factors, the neural network structure or hyperparameters were adjusted and retrained until the trained neural network could effectively capture the dynamic relationship between liquid film thickness and environmental factors. The trained neural network effectively captures the dynamic relationship between liquid film thickness and environmental factors, thus completing the training and accuracy verification of the neural network.

[0015] In a specific embodiment of the present invention, the calculation formula for the accuracy index is as follows:

[0016] in, For accuracy indicators, To test the true values ​​of the sample set; is the predicted value; n is the number of samples.

[0017] A second objective of this invention is to provide a real-time liquid film thickness prediction method, comprising: Based on real-time environmental factor data and the aforementioned correlation model between corrosion environmental factors and liquid film thickness, the real-time liquid film thickness is predicted.

[0018] In a specific embodiment of the present invention, the real-time liquid film thickness prediction method further includes: The corrosion morphology of the metal surface is determined based on the predicted liquid film thickness.

[0019] The third objective of this invention is to provide a system for constructing a correlation model between corrosion environmental factors and liquid film thickness, comprising: Calculation module: used to obtain liquid film thickness based on corrosion environment factor data of the target area; Training and Validation Module: Used to construct a dataset with corrosion environment factor data as input features and liquid film thickness as output feature; also used to train and validate the accuracy of the neural network based on the dataset; The acquisition module is used to use the neural network after training and accuracy verification as a correlation model between corrosion environment factors and liquid film thickness.

[0020] In a specific embodiment of the present invention, the calculation module includes a salt spray deposition rate submodule and a liquid film thickness submodule; The salt spray deposition rate submodule is used to calculate the salt spray deposition rate based on the chloride ion content in the corrosion environment factor data; The liquid film thickness submodule is used to obtain the liquid film thickness based on the relative humidity and salt spray deposition rate in the corrosion environment factor data.

[0021] The beneficial effects of this invention are: This invention discloses a method and system for constructing a correlation model between corrosion environmental factors and liquid film thickness. By proposing to use the liquid film thickness on the metal surface as an intermediate key parameter and as the core driving factor of the corrosion occurrence and development mechanism, and constructing a correlation model of environmental factors around the liquid film thickness, the invention achieves the real-time acquisition of liquid film thickness, that is, the dynamic prediction of liquid film thickness.

[0022] The corrosion environment factor and liquid film thickness correlation model of the present invention is a neural network-based prediction model and a nonlinear prediction model. It comprehensively considers the coupling effect of factors such as temperature and humidity, corrosive medium concentration and air pressure on the liquid film thickness on the metal surface. It effectively overcomes the defects of traditional models in complex environments, such as low accuracy and poor adaptability, improves the scientific nature of liquid film modeling, and provides key support for subsequent corrosion assessment, protection design and life prediction.

[0023] This invention employs correlation coefficient method / sensitivity analysis to screen out key variables that significantly affect liquid film thickness from multiple corrosion environmental factors, including relative humidity, temperature, atmospheric pressure, chloride ion content, wind speed, wind direction, distance from coastline, altitude, ammonia concentration, sulfur dioxide concentration, and nitrogen dioxide concentration. These variables are then used as model inputs, thereby improving the scientific rigor and accuracy of the modeling.

[0024] This invention addresses the nonlinear dynamic changes in liquid film thickness with environmental factors by introducing a Long Short-Term Memory (LSTM) neural network to construct a correlation model between corrosion environmental factors and liquid film thickness. The model structure includes two LSTM hidden layers and a Dropout layer to prevent overfitting. This model possesses strong historical dependency modeling and trend prediction capabilities, and can adapt to the dynamic evolution of liquid film thickness under complex environments.

[0025] Compared with existing liquid film thickness measurement techniques based on optical measurements, the method of this invention does not rely on complex experimental equipment and is suitable for continuous deployment in natural environments. Unlike modeling methods that only focus on corrosion rate or outcome variables, this invention uses liquid film thickness as the core variable to reflect the precursory physical processes of corrosion, enhancing the physical interpretability of the model and the clarity of its predictive logic. Furthermore, by combining LSTM neural network modeling, it can integrate multi-source corrosion environmental factors, effectively modeling the nonlinear coupling and temporal dynamic evolution of environmental factors such as relative humidity, temperature, atmospheric pressure, chloride ion content, wind speed, wind direction, distance from the coast, altitude, ammonia concentration, sulfur dioxide concentration, and nitrogen dioxide concentration with liquid film thickness. This overcomes the linear assumptions of traditional statistical models and possesses greater adaptability and engineering application value.

[0026] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A flowchart of a method for constructing a correlation model between corrosion environment factors and liquid film thickness according to an embodiment of the present invention is shown; Figure 2 A correlation analysis graph of liquid film thickness according to an embodiment of the present invention is shown; Figure 3 A schematic diagram of a neural network structure according to an embodiment of the present invention is shown; Figure 4 A comparison chart of the actual and predicted values ​​according to an embodiment of the present invention is shown; Figure 5 A distribution curve of accuracy indicators according to an embodiment of the present invention is shown; Figure 6 A framework diagram of a system for constructing a correlation model between corrosion environment factors and liquid film thickness according to an embodiment of the present invention is shown; In the diagram: Calculation module 10; Training and validation module 20; Acquisition module 30. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] like Figure 1 As shown, a method for constructing a correlation model between corrosion environment factors and liquid film thickness according to certain embodiments of the present invention includes: S1. Obtain the liquid film thickness based on the corrosion environment factor data of the target area; S2. Construct a dataset with corrosion environment factor data as input features and liquid film thickness as output feature; S3. Based on the dataset, train and verify the accuracy of the neural network; S4. The neural network after training and accuracy verification is used as a correlation model between corrosion environment factors and liquid film thickness.

[0031] In some embodiments of the present invention, in step S1, the corrosive environmental factors include relative humidity, temperature, atmospheric pressure, chloride ion content, wind speed, wind direction, distance from the coast, altitude, ammonia concentration, sulfur dioxide concentration, and nitrogen dioxide concentration.

[0032] In some embodiments of the present invention, the real-time corrosion environmental factor data is collected by a corrosion environment observation station to collect environmental parameters of the target area (such as tropical islands and reefs).

[0033] In some embodiments of the present invention, step S1 includes: S1-1. Calculate the salt spray deposition rate based on the chloride ion content in the corrosion environment factor data; S1-2. Based on the relative humidity and salt spray deposition rate in the corrosion environment factor data, the liquid film thickness is obtained.

[0034] In some embodiments of the present invention, in step S1-1, the formula for calculating the salt spray deposition rate is as shown in equation (1): (1) In equation (1), DD Salt spray deposition rate, c salt Chloride ion content, v f The correlation factor is calculated using the formula (2): (2) In equation (2), g represents the acceleration due to gravity, and η is the dynamic viscosity of the salt spray. Indicates salt spray density, Indicates air density, r The salt spray particle size is expressed in cm. Both the salt spray particle size and salt spray concentration distribution can be obtained through on-site measurements. When the salt spray particle size is difficult to measure, it can be calculated using formula (3): (3) In equation (3), c 1 is a constant, 0.7674; c 2 is a constant, 3.079; c 3 is a constant, 2.573 × 10 -11 ; c 4 is a constant, -1.424; r d The particle size of the dried salt nuclei is expressed in centimeters (cm). r d =5×10 -5 Substitute cm into equation (3) to calculate the salt spray particle size.

[0035] In some embodiments of the present invention, step S1-2 includes: S1-2-1. Determine whether the relative humidity of the environment is greater than the critical humidity of the metal surface; S1-2-2. Based on the judgment that the relative humidity of the environment is greater than the critical humidity of the metal surface (that is, the environment has reached the conditions for the formation of liquid film on the metal surface), the thickness of the liquid film is calculated according to the salt spray deposition rate and the relative humidity of the environment. Specifically, the formula for calculating the thickness of the liquid film is shown in equation (4): (4) In equation (4), FT For liquid film thickness, DD Salt spray deposition rate, RH Relative humidity, a 1. a 2. a 3. b 1. b 2. b 3 are all constants.

[0036] In some embodiments of the present invention, step S2, which uses corrosion environment factor data as input features, includes: i. Perform correlation analysis on liquid film thickness using corrosion environmental factor data; ii. Based on the correlation analysis results of liquid film thickness, key corrosion environment factor data were selected; iii. Use key corrosion environment factor data as input features.

[0037] In some embodiments of the present invention, in step S3, the neural network is a Long Short-Term Memory (LSTM) neural network, which includes an input layer, two hidden layers, and an output layer, used to fit the nonlinear temporal mapping relationship between input environmental factors and liquid film thickness. The first LSTM hidden layer contains 64 neurons for extracting short-term temporal features; the second LSTM hidden layer contains 32 neurons to further capture long-term dependency information; a Dropout layer is connected after the hidden layers with a dropout rate set to 0.2 to prevent overfitting; the output layer has one neuron for regressing and predicting the liquid film thickness value. The LSTM neural network uses mean squared error (MSE) as the loss function, and the Adam optimizer is selected with an initial learning rate set to 0.001.

[0038] In some embodiments of the present invention, step S4 includes: S4-1. Train the neural network based on the training sample set in the dataset; S4-2. Validate the trained neural network based on the test sample set in the dataset; S4-3. Calculate the accuracy index data based on the verification results; S4-4. Based on the accuracy index data, determine whether the trained neural network effectively captures the dynamic relationship between liquid film thickness and environmental factors. S4-5. Since the neural network after training failed to effectively capture the dynamic relationship between liquid film thickness and environmental factors, the neural network structure or hyperparameters were adjusted and retrained until the neural network after training effectively captured the dynamic relationship between liquid film thickness and environmental factors. S4-6. Based on the completed training of the neural network, effectively capture the dynamic relationship between the liquid film thickness and environmental factors, and complete the training and accuracy verification of the neural network.

[0039] In some embodiments of the present invention, in step S4-3, the calculation formula for the accuracy index is as shown in equation (5): (5) in, For accuracy indicators, To test the true values ​​of the sample set; is the predicted value; n is the number of samples.

[0040] In some embodiments of the present invention, step S4-4 includes: Based on the accuracy index data, draw the distribution curve of the accuracy index (i.e., visualize the prediction results of the validation set based on the accuracy index data). Observe whether the distribution curve of the accuracy index shows any significant deviation or time-series drift in its trend; If the distribution curve of the accuracy index shows a significant deviation in trend or a time-series drift, it is determined that the neural network after training has not effectively captured the dynamic relationship between the liquid film thickness and environmental factors. If the distribution curve of the accuracy index does not show a significant deviation or temporal drift in trend, then the neural network after training is deemed to effectively capture the dynamic relationship between liquid film thickness and environmental factors.

[0041] In some embodiments of the present invention, a significant deviation or timing drift is considered to exist when the accuracy index exceeds 30%.

[0042] A real-time liquid film thickness prediction method according to certain embodiments of the present invention includes: Based on real-time environmental factor data and the corrosion environment factor and liquid film thickness correlation model in the above embodiments, the real-time liquid film thickness is predicted. Specifically, the real-time environmental factor data is input into the corrosion environment factor and liquid film thickness correlation model, and the predicted real-time liquid film thickness is obtained through the corrosion environment factor and liquid film thickness correlation model.

[0043] In some embodiments of the present invention, the prediction method further includes: The corrosion morphology of the metal surface is determined based on the predicted liquid film thickness.

[0044] The corrosion morphology of a metal surface is determined based on the magnitude of the liquid film thickness, for example including: FT <10nm, the corrosion morphology of the metal surface is determined to be dry atmospheric corrosion; 10nm≤ FT If the value is less than 1 μm, the corrosion morphology of the metal surface is determined to be atmospheric corrosion. 1um≤ FT If the corrosion rate is ≤1mm, the corrosion morphology of the metal surface is determined to be humid atmospheric corrosion. FT >1mm, indicating a liquid corrosion or salt water immersion environment.

[0045] A correlation model between corrosion environment factors and liquid film thickness was constructed using the aforementioned method.

[0046] The thickness of the liquid film is obtained by calculating using equations (1)-(4).

[0047] In step i, a correlation analysis of liquid film thickness is performed on the corrosion environmental factor data to screen out environmental factors with a correlation coefficient greater than 0.05, including humidity, temperature, atmospheric pressure, chloride ion content, SO2, NO2, and NH3 content, such as... Figure 2 As shown, the above environmental factor data are used as input features, and the real-time liquid film thickness is used as the output feature to construct a dataset.

[0048] The LSTM network structure in this example is as follows: Figure 3 As shown.

[0049] The dataset is divided into a training sample set and a test sample set, for example, in a ratio of 7:3.

[0050] Based on the training sample set Figure 3 The LSTM network was trained using mean squared error (MSE) as the loss function and Adam optimizer as the optimizer. The initial learning rate was set to 0.001.

[0051] After training, accuracy verification is performed: the predicted curve values ​​obtained through the neural network are compared with the true value curves, such as... Figure 4 As shown; Further calculate the accuracy index data according to equation (5), and further plot the distribution curve of the accuracy index, as shown in the figure. Figure 5 As shown, accuracy indicators r MAPE There is no significant deviation or time-series drift in the trend, and the average value over the entire prediction period is only 8.66%, indicating that the model can effectively capture the dynamic relationship between liquid film thickness and environmental factors.

[0052] The neural network, trained and validated for accuracy, serves as a model relating corrosion environmental factors and liquid film thickness.

[0053] The real-time liquid film thickness is predicted using a correlation model between corrosion environmental factors and liquid film thickness, along with real-time environmental factor data.

[0054] The corrosion morphology of the metal surface is determined by the magnitude of the predicted liquid film thickness.

[0055] The aforementioned correlation model underwent preliminary feasibility verification through comparison with historical measured data and numerical simulation. Multi-source meteorological and corrosion factor data from typical highly corrosive environments (such as tropical island and reef regions) were selected, and sample training and prediction backtesting were conducted using liquid film formation theory. The model's prediction results showed a high degree of consistency with the measured liquid film thickness variation trends reported in the literature, demonstrating good time-series response capability and prediction accuracy. The root mean square error (RMSE) was only 8.66%, indicating that the model possesses strong generalization ability and engineering adaptability, laying the foundation for subsequent experimental verification and engineering deployment.

[0056] like Figure 6 As shown, a corrosion environment factor and liquid film thickness correlation model construction system according to certain embodiments of the present invention includes: Calculation module 10: used to obtain the liquid film thickness based on corrosion environment factor data of the target area; Training and Validation Module 20: This module is used to construct a dataset with corrosion environment factor data as input features and liquid film thickness as output features; it is also used to train and validate the accuracy of the neural network based on the dataset. Module 30: Used to use the neural network after training and accuracy verification as a correlation model between corrosion environment factors and liquid film thickness.

[0057] In some embodiments of the present invention, the calculation module 10 includes a salt spray deposition rate submodule and a liquid film thickness submodule; The salt spray deposition rate submodule is used to calculate the salt spray deposition rate based on the chloride ion content in the corrosion environment factor data; The liquid film thickness submodule is used to obtain the liquid film thickness based on the relative humidity and salt spray deposition rate in the corrosion environment factor data.

[0058] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a correlation model between corrosion environmental factors and liquid film thickness, characterized in that, include: The liquid film thickness is obtained based on corrosion environment factor data of the target area; Construct a dataset with corrosion environment factor data as input features and liquid film thickness as output feature; Based on the dataset, the neural network is trained and its accuracy is verified. The neural network, after training and accuracy verification, was used as a correlation model between corrosion environment factors and liquid film thickness.

2. The method for constructing a correlation model between corrosion environment factors and liquid film thickness according to claim 1, characterized in that, The corrosive environmental factors include relative humidity, temperature, atmospheric pressure, chloride ion content, wind speed, wind direction, distance from the coast, altitude, ammonia concentration, sulfur dioxide concentration, and nitrogen dioxide concentration.

3. The method for constructing a correlation model between corrosion environment factors and liquid film thickness according to claim 1, characterized in that, The liquid film thickness is obtained based on the corrosion environment factor data of the target area, including: Calculate the salt spray deposition rate based on the chloride ion content in the corrosion environment factor data; The liquid film thickness was obtained based on the relative humidity and salt spray deposition rate from the corrosion environment factor data.

4. The method for constructing a correlation model between corrosion environmental factors and liquid film thickness according to claim 3, characterized in that, The method of obtaining liquid film thickness based on relative humidity and salt spray deposition rate from corrosion environment factor data includes: Determine whether the relative humidity of the environment is greater than the critical humidity of the metal surface; Based on the judgment that the relative humidity of the environment is greater than the critical humidity of the metal surface, the liquid film thickness is calculated according to the salt spray deposition rate and the relative humidity of the environment.

5. The method for constructing a correlation model between corrosion environment factors and liquid film thickness according to claim 4, characterized in that, The formula for calculating the thickness of the liquid film is as follows: in, FT For liquid film thickness, DD Salt spray deposition rate, RH Relative humidity, a 1. a 2. a 3. b 1. b 2. b 3 are all constants.

6. The method for constructing a correlation model between corrosion environment factors and liquid film thickness according to claim 1, characterized in that, The method of using corrosion environment factor data as input features includes: Correlation analysis of liquid film thickness was performed on corrosion environmental factor data; Based on the correlation analysis results of liquid film thickness, key corrosion environment factor data were selected; Key corrosion environmental factor data were used as input features.

7. The method for constructing a correlation model between corrosion environmental factors and liquid film thickness according to claim 1, characterized in that, The neural network is a long short-term memory neural network, which includes an input layer, two hidden layers and an output layer, and is used to fit the nonlinear temporal mapping relationship between input environmental factors and liquid film thickness.

8. A method for constructing a correlation model between corrosion environmental factors and liquid film thickness according to any one of claims 1-7, characterized in that, The training and accuracy verification of the neural network based on the dataset includes: The neural network is trained based on the training sample set in the dataset; The trained neural network is validated based on the test sample set in the dataset; Based on the verification results, calculate the accuracy index data; Based on the accuracy index data, determine whether the trained neural network effectively captures the dynamic relationship between liquid film thickness and environmental factors. Since the trained neural network failed to effectively capture the dynamic relationship between liquid film thickness and environmental factors, the neural network structure or hyperparameters were adjusted and retrained until the trained neural network could effectively capture the dynamic relationship between liquid film thickness and environmental factors. The trained neural network effectively captures the dynamic relationship between liquid film thickness and environmental factors, thus completing the training and accuracy verification of the neural network.

9. The method for constructing a correlation model between corrosion environmental factors and liquid film thickness according to claim 8, characterized in that, The formula for calculating the accuracy index is as follows: in, For accuracy indicators, To test the true values ​​of the sample set; is the predicted value; n is the number of samples.

10. A method for real-time prediction of liquid film thickness, characterized in that, include: Based on real-time environmental factor data and a correlation model between corrosion environmental factors and liquid film thickness as described in any one of claims 1-9, the real-time liquid film thickness is predicted.

11. The real-time liquid film thickness prediction method according to claim 10, further comprising: The corrosion morphology of the metal surface is determined based on the predicted liquid film thickness.

12. A system for constructing a correlation model between corrosion environmental factors and liquid film thickness, characterized in that, include: Calculation module: used to obtain liquid film thickness based on corrosion environment factor data of the target area; Training and validation module: used to build a dataset with corrosion environment factor data as input features and liquid film thickness as output feature; It is also used for training and accuracy verification of neural networks based on datasets; The acquisition module is used to use the neural network after training and accuracy verification as a correlation model between corrosion environment factors and liquid film thickness.

13. The system for constructing a correlation model between corrosion environmental factors and liquid film thickness according to claim 12, characterized in that, The calculation module includes a salt spray deposition rate submodule and a liquid film thickness submodule; The salt spray deposition rate submodule is used to calculate the salt spray deposition rate based on the chloride ion content in the corrosion environment factor data; The liquid film thickness submodule is used to obtain the liquid film thickness based on the relative humidity and salt spray deposition rate in the corrosion environment factor data.

Images