Epoxy coating degradation prediction method considering multi-environment comprehensive influence shielding effect

Abstract

The invention discloses an epoxy coating degradation prediction method considering a multi-environment comprehensive influence shielding effect, and belongs to the technical field of epoxy coating environmental adaptability assessment, and the method comprises the steps: 1, designing an epoxy coating performance degradation test, and collecting epoxy coating environmental factor data and degradation information data; step 2, after the data in the step 1 are obtained, establishing an environmental effect model by using a random forest model optimized based on a particle swarm algorithm, and training the environmental effect model; 3, establishing a natural environment data model according to the external field service natural environment data of the epoxy coating; 4, performing degradation prediction on the coating performance in an external field service environment; according to the epoxy coating degradation prediction method considering the multi-environment comprehensive influence shielding effect provided by the invention, the problem that the nonlinear degradation of the epoxy coating is difficult to predict due to the complexity of natural environment factors is solved, and the accuracy and reliability of the prediction of the degradation rate of the epoxy coating are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental adaptability assessment of epoxy coatings, and in particular to a method for predicting epoxy coating degradation under the shielding effect of multiple environmental comprehensive influences. Background Art

[0002] Epoxy coatings are widely used for corrosion protection of industrial and civilian metal structures. However, in service environments, epoxy coatings gradually degrade. Failure to accurately assess and predict the corrosion protection performance of coatings can lead to economic losses and safety risks. The performance of epoxy coatings in outdoor environments is significantly affected by environmental factors, with UV index, temperature, and humidity being the primary factors contributing to degradation. Degradation products from epoxy coatings adhere to the coating surface, weakening the effects of the natural environment on coating degradation. Existing degradation has a direct and significant impact on subsequent degradation processes, resulting in a nonlinear degradation trajectory, where the degradation rate decreases as the amount of degradation increases. This phenomenon can be attributed to the shielding effect of the degradation products.

[0003] Traditional methods for predicting material degradation in service environments are primarily based on mathematical tools such as degradation path fitting and multivariate regression. These methods construct mathematical mappings between material performance characteristic parameters, environmental characteristic parameters, and degradation time. However, actual service environments are often highly complex and dynamically variable, resulting in poor accuracy and generalization of prediction models. In recent years, machine learning technology has demonstrated a powerful ability to process complex data in applications across multiple fields. It can automatically learn patterns from complex, multidimensional datasets and capture the nonlinear characteristics of material degradation processes and the interactions between key variables. Therefore, it has enormous application potential in the field of material degradation prediction. In machine learning, the random forest model optimized by the particle swarm algorithm combines the optimization properties of the particle swarm with the strong generalization capabilities of the random forest algorithm, further improving classification accuracy and model generalization.

[0004] Based on this, the present invention proposes a method for predicting epoxy coating degradation under the shielding effect of multiple environments. It comprehensively considers the influence of different factors in the degradation process of epoxy coating, combines the time series prediction method and the random forest model optimized by particle swarm algorithm to give a method for predicting the degradation rate of epoxy coating. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for predicting epoxy coating degradation under the shielding effect of multiple environmental comprehensive influences, so as to solve the problem in the above-mentioned background technology that the nonlinear degradation of epoxy coating caused by the complexity of natural environmental factors is difficult to predict.

[0006] To achieve the above object, the present invention provides a method for predicting epoxy coating degradation under the shielding effect of multiple environmental comprehensive influences, comprising the following steps:

[0007] Step 1: Design an epoxy coating performance degradation test and collect epoxy coating environmental factor data and degradation information data;

[0008] Step 2: After obtaining the data from step 1, a random forest model optimized by particle swarm optimization is used to establish a coating performance prediction model and conduct training;

[0009] Step 3: Establishing an environmental data model based on the field service environmental data of the epoxy coating;

[0010] Step 4: Predict the degradation of coating performance under field service environment.

[0011] Preferably, step one is specifically:

[0012] Epoxy resin coatings are prepared into test samples, which are then placed in batches at different time points in an outdoor natural exposure test field or a laboratory simulated field environment to conduct coating degradation tests, controlling the different environments in which the test samples are exposed. During the test, UV intensity sensors, temperature sensors, and relative humidity sensors are placed to automatically and regularly record environmental factor data such as UV coefficient, temperature, and humidity, as well as the time data during the degradation process.

[0013] During the test, the degradation performance of the coating test samples is monitored in real time, such as gloss, color difference, electrochemical impedance or the number of functional group structures of specific compounds, and the cumulative degradation amount of the coating degradation performance corresponding to different test times is recorded. By analyzing the degradation performance of the test samples in the degradation test, the degradation process can be effectively quantified, and the degradation information of the epoxy coating under the influence of different environmental factors and degradation products can be obtained.

[0014] Preferably, in step 2, after obtaining the performance degradation data and environmental data of the test samples under different environments, a random forest model optimized by a particle swarm algorithm is used to establish an environmental effect model. During the degradation process of the epoxy coating under a service environment, the amount of degradation that has already occurred will have a direct and significant impact on the subsequent degradation process. Therefore, when training the environmental effect model, the nonlinearity of material degradation is taken into account, and the UV coefficient, temperature, relative humidity, degradation time, and cumulative degradation amount under different environments are selected as input features to train the random forest model optimized by the particle swarm algorithm. This will result in a model that can predict the degradation state of the epoxy coating under service conditions.

[0015] The particle swarm optimization algorithm uses its global search capability in continuous optimization problems to optimize the model's hyperparameters, where the set of candidate solutions to the optimization problem is defined as a particle swarm to find the optimal solution:

[0016]

[0017] Among them, X i =(x i1 ,x i2 ,x i3 ,......,x iD ) space vector is represented as the i-th particle, V i =(v i1 ,v i2 ,v i3 ,......,v iD ) is the flying speed of the i-th particle, P i =(p i1 ,p i2 ,p i3 ,......,p iD ) is the optimal position of the particle; ω is the inertia weight; c1 and c2 are learning factors; r1 and r2 are random numbers between [0,1]; p gd is the current position of the particle; p id 、 The subscript i in is [i,m], where m is the particle swarm, and the subscript d is [d,D], where D is the number of iterations.

[0018] The random forest model introduces the idea of random feature space on the basis of decision trees to ensure the diversity of each decision subtree and enhance the algorithm's anti-noise performance. The final prediction result is the average of the prediction values of each tree:

[0019]

[0020] Among them, Y(x) represents the overall prediction result; h(x,α n ) represents the output based on x and α, T represents the number of decision trees in the random forest, n represents the number of features considered when each tree is built, x represents the input value of the nth tree, and α represents the parameter of the nth tree.

[0021] Preferably, step three specifically includes:

[0022] The monthly average data of UV coefficient, temperature, and relative humidity in natural environments were collected from the public natural environment database, and a natural environment data model was established. In actual natural environments, UV intensity, temperature, and relative humidity show dynamic periodic changes and random fluctuation characteristics, usually with a cycle of one year. In order to accurately describe the periodic fluctuation and random fluctuation characteristics of environmental factors in the natural environment, a sinusoidal time-varying function is used to describe the dynamic periodicity of the environmental factor changes, and a random term is added to describe its random fluctuation, which is:

[0023]

[0024] Among them, ψ(t)=[UV(t),T(t),RH(t)] T Used to describe the natural environment ultraviolet intensity, temperature and relative humidity at time t, α0 = [α UV0 ,α T0 ,α RH0 ] T is a constant used to describe the annual average level of ultraviolet intensity, temperature and relative humidity in the natural environment, α1=[α UV1 ,α T1 ,α RH1 ] T Used to describe the upper and lower limits of ultraviolet intensity, temperature and relative humidity in the natural environment. Used to determine the time when the ultraviolet intensity, temperature and relative humidity in the natural environment reach the extreme value during periodic changes, ε=[ε UV ,ε T ,ε RH ] T Used to describe the random fluctuations of UV intensity, temperature and relative humidity in the natural environment, obeying the multivariate normal distribution MN(μ,Σ):

[0025] μ=[0,0,0] T ;

[0026]

[0027] Based on the collected natural environment data, the maximum likelihood estimation method is used to estimate the above parameters.

[0028] Preferably, the basic idea of the maximum likelihood estimation method is to maximize the likelihood function (probability) of the data observed under these parameters by selecting a set of parameters. The likelihood function is expressed as the product of the probabilities of each sample, which is:

[0029]

[0030] Its goal is to find the parameter θ so that the likelihood function L(θ) reaches its maximum value.

[0031] Preferably, after obtaining the parameter estimation value by the maximum likelihood estimation method in step 4, the Monte Carlo simulation method is used to predict the degradation state of the epoxy coating under the field service environment, specifically: randomly generate a large number of multivariate normal distribution random numbers according to the results of the parameter estimation, generate a large number of natural environment simulation data sequences, input the natural environment simulation data sequences into the trained random forest model based on particle swarm algorithm optimization, and output the degradation state of the epoxy coating in the natural environment; determine the degradation interval according to the quantile of the degradation amount at different degradation moments, and obtain the interval prediction result of the degradation state of the epoxy coating, and obtain the point estimate of the mean value of the degradation state at different times according to the 50% quantile, so as to finally achieve accurate prediction of the degradation state of the epoxy coating under the field service environment.

[0032] Therefore, the present invention adopts the above-mentioned epoxy coating degradation prediction method considering the shielding effect of multiple environments, which has the following beneficial effects:

[0033] (1) Aiming at the nonlinear degradation problem of epoxy coatings under natural environment, the shielding effect of the adhesion of the degradation products produced will have a direct and significant impact on the subsequent degradation process, thus ensuring the generalization of the degradation state prediction of epoxy coatings;

[0034] (2) When training the coating performance prediction model based on the random forest algorithm, UV intensity, temperature, relative humidity, degradation time and cumulative degradation amount are used as input features, and the performance degradation of the coating is used as the output feature, so that the model prediction results are more accurate;

[0035] (3) The method used has low requirements for the amount of environmental data, is quick and simple to solve, and has strong operability. The evaluation results are more stable, which reflects its superiority in processing small sample data.

[0036] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of the epoxy coating degradation prediction method considering the shielding effect of multiple environments;

[0038] Figure 2 A schematic diagram comparing the degradation paths of the predicted and true values of the test set results reserved for the embodiment of the present invention;

[0039] Figure 3 This is the epoxy coating status prediction result under the field service environment of the embodiment of the present invention. DETAILED DESCRIPTION

[0040] The following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort shall fall within the scope of protection of the present invention.

[0041] In this embodiment, a natural environment data collection experiment for a certain epoxy coating was conducted in an outdoor environmental laboratory from 2002 to 2006, with a total of 36 samples placed. The exposure time of these samples was different, and they were exposed in batches within a time span of about five years. During the experiment, the sensor automatically recorded environmental data such as temperature and humidity, and the ultraviolet coefficient data referred to the ultraviolet index of the local exposure day provided by the Meteorological Bureau. The degradation state of the sample is represented by the number of functional groups of a specific compound. The peak generated by the specific functional group structure in the sample on the Fourier infrared spectrum has a height that is proportional to the concentration of the compound or structure. Therefore, the degradation process is quantified by analyzing the change in the height of the specific peak on the Fourier infrared spectrum. Assuming that the degradation mechanism of the data collected in the experiment is consistent with the degradation mechanism in the service environment, a natural environment and environmental effect model is established according to a method for predicting epoxy coating degradation under the shielding effect of multiple environmental comprehensive influences proposed by the present invention, and the degradation state of the epoxy coating is predicted.

[0042] In this example, the following basic settings need to be established:

[0043] (1) For epoxy coatings of the same material, the degradation mechanism when it degrades in the test environment is consistent with the degradation mechanism when it degrades in the actual use environment;

[0044] (2) When considering the degradation process of epoxy coatings, the main factors to be considered are the UV coefficient, temperature, and humidity, which have the greatest impact on the degradation rate of epoxy coatings in the natural environment.

[0045] See also Figure 1 A method for predicting epoxy coating degradation under shielding effects of multiple environments is proposed, comprising the following steps:

[0046] Step 1: Epoxy coating performance degradation test design and data collection

[0047] First, epoxy resin coatings were prepared into test samples. These samples were then placed in batches at different time points in an outdoor, natural environment on a rooftop to conduct coating degradation tests. During the tests, temperature and relative humidity sensors automatically recorded environmental data at regular intervals.

[0048] During the test, the height of the peak at 1250 cm-1 on the infrared spectrum of the coating test sample was monitored in real time, which is a representation of the number of certain functional groups of the aryl ether. The collected experimental data are shown in Table 1:

[0049] Table 1 Partial natural environment degradation test data of an epoxy coating

[0050]

[0051]

[0052] Step 2: Training of coating performance prediction model based on random forest algorithm

[0053] After obtaining performance degradation data and environmental data from test samples under different environments, an environmental effect model was constructed using a random forest model optimized with a particle swarm algorithm. During training, environmental impact factors and various degradation information were used as input variables, and random sampling was used to shuffle the data. From the constructed dataset, 80% of the data was used to form the training set, which was used to train and build a highly accurate prediction model. The remaining 20% of the data was used to form the validation set, which was used to adjust the model's hyperparameters. In addition, five sets of data were reserved as test sets to evaluate the training results and model generalization.

[0054] After the model training is completed, the five reserved test sets are input into the model to evaluate the training effect. The true value of the model is compared with the predicted value. The evaluation indicators of the model accuracy are shown in Table 2:

[0055] Table 2. Model accuracy evaluation results after training

[0056] Rating indicators result Goodness of fit of training set 0.9430 Test set goodness of fit 0.8769 Mean absolute error of the training set 0.0022 The mean absolute error of the test set 0.0034 Root mean square error of the training set 0.0053 Root mean square error of the test set 0.00073

[0057] The comparison between the predicted value and the true value degradation path of the reserved test set result is shown in Figure 2 .

[0058] Step 3: Modeling the coating's field service environment

[0059] We collected 12-month monthly mean data for temperature and relative humidity in a city from the China Meteorological Administration's public data website, and 12-month monthly mean data for the UV index in a city from the spheric emission monitoring internet service website. The data results are shown in Table 3.

[0060] Table 3 Results of collected monthly average data in Beijing in a certain year

[0061] month Temperature / ℃ Relative humidity / % UV coefficient 1 -3 42 1.467387 2 0 43 2.459179 3 8 38 4.092742 4 15 41 6.3483 5 22 43 7.993613 6 26 56 8.267267 7 28 68 9.211258 8 27 68 8.325097 9 21 62 6.406433 10 14 59 3.924323 11 5 54 2.131167 12 -1 43 1.43271

[0062] Based on the data in Table 3, the maximum likelihood method is used to estimate the parameters of the natural environment model. The results are as follows:

[0063]

[0064] Step 4: Prediction of coating performance degradation under field service environment

[0065] After obtaining the aforementioned parameter estimates using a maximum likelihood estimation algorithm, the random numbers were processed based on the parameter estimation results to generate 1,000 multivariate normally distributed random numbers. These random numbers were then substituted into the natural environment model to construct a large number of simulated natural environment data sequences. Through the natural environment model, environmental factor data for the epoxy coating at different degradation points during field service can be output.

[0066] The simulation data sequence obtained by extrapolating the natural environment model is input into the prediction model that has been trained. In order to further quantify the predicted degradation state of the epoxy coating, the 95% interval of the monthly degradation state data is divided as the degradation interval of the prediction model according to the degradation quantity quantiles of different months, and the interval prediction results of the coating degradation state are obtained; the 50% quantile is used to extract the mean of the degradation state of different months as the point estimate of the prediction result, which more intuitively reflects the overall trend of the degradation process, and finally realizes the accurate prediction of the degradation state of the epoxy coating in the field service environment. The prediction results of the epoxy coating state in the field service environment collected are as follows: Figure 3 shown.

[0067] Therefore, the present invention adopts the above-mentioned epoxy coating degradation prediction method under the shielding effect of multiple environmental comprehensive influences. Aiming at the problem that the complexity of natural environmental factors makes it difficult to predict the nonlinear degradation of epoxy coatings, the present invention designs experiments to collect environmental data of epoxy coatings for training a random forest model optimized based on a particle swarm algorithm, establishes a natural environment model to quantify the periodic changes and uncertainties of environmental factors, uses the Monte Carlo method to reasonably extrapolate the data, and inputs the obtained simulation data sequence as a feature into the trained prediction model to predict the degradation state zone of the epoxy coating under field service environment. The method is suitable for fields such as environmental adaptability assessment of epoxy coatings in natural environments and has strong operability.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solutions of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for predicting epoxy coating degradation under the shielding effect of multiple environments, characterized by: The following steps are involved: Step 1: Design an epoxy coating performance degradation test and collect epoxy coating environmental factor data and degradation information data; Step 2: After obtaining the data from step 1, a random forest model optimized by particle swarm optimization is used to establish a coating performance prediction model and conduct training; Step 3: Establishing an environmental data model based on the field service environmental data of the epoxy coating; Step 4: Predict the degradation of coating performance under field service environment.

2. The epoxy coating degradation prediction method considering the shielding effect of multiple environments according to claim 1 is characterized in that: Step 1 is as follows: Epoxy resin coatings are prepared into test samples, which are then placed in batches at different time points in an outdoor natural exposure test field or a laboratory simulated field environment to conduct coating degradation tests, controlling the different environments in which the test samples are exposed. During the test, UV intensity sensors, temperature sensors, and relative humidity sensors are placed to automatically and regularly record environmental factor data such as UV coefficient, temperature, and humidity, as well as the time data during the degradation process. During the test, the degradation performance of the coating test samples was monitored in real time, and the cumulative degradation amount of the coating degradation performance corresponding to different test times was recorded. By analyzing the degradation performance of the test samples in the degradation test, the degradation information of the epoxy coating under the influence of different environmental factors and degradation products was obtained.

3. The epoxy coating degradation prediction method considering the shielding effect of multiple environments according to claim 2 is characterized in that: Step 2: After obtaining the data from step 1, an environmental effect model is established using a random forest model optimized by a particle swarm algorithm. The nonlinearity of material degradation performance is taken into account when training the environmental effect model. The UV coefficient, temperature, relative humidity, degradation time, and cumulative degradation amount under different environments are selected as input features to train the random forest model optimized by the particle swarm algorithm, thereby obtaining a model for predicting the degradation state of the epoxy coating under service conditions. The particle swarm optimization algorithm uses its global search capability in continuous optimization problems to optimize the model's hyperparameters, where the set of candidate solutions to the optimization problem is defined as a particle swarm to find the optimal solution: Among them, X i =(x i1 ,x i2 ,x i3 ,......,x iD ) space vector is represented as the i-th particle, V i =(v i1 ,v i2 ,v i3 ,......,v iD ) is the flying speed of the i-th particle, P i =(p i1 ,p i2 ,p i3 ,......,p iD ) is the optimal position of the particle; ω is the inertia weight; c1 and c2 are learning factors; r1 and r2 are random numbers between [0,1]; p gd is the current position of the particle; p id 、 The subscript i in is [i,m], where m is the particle swarm, and the subscript d is [d,D], where D is the number of iterations. The random forest model introduces the idea of random feature space based on the decision tree. The final prediction result is the average of the prediction values of each tree: Among them, Y(x) represents the overall prediction result; h(x,α n ) represents the output based on x and α, T represents the number of decision trees in the random forest, n represents the number of features considered when each tree is built, x represents the input value of the nth tree, and α represents the parameter of the nth tree.

4. The epoxy coating degradation prediction method considering the shielding effect of multiple environments according to claim 3 is characterized in that: Step three specifically includes: The monthly mean data of UV coefficient, temperature, and relative humidity in the natural environment are collected from the public natural environment database, and a natural environment data model is established. A sinusoidal time-varying function is used to describe the dynamic periodicity of changes in environmental factors, and a random term is added to describe its random fluctuations, which is: Among them, ψ(t)=[UV(t),T(t),RH(t)] T Used to describe the natural environment ultraviolet intensity, temperature and relative humidity at time t, α0 = [α UV0 ,α T0 ,α RH0 ] T is a constant used to describe the annual average level of ultraviolet intensity, temperature and relative humidity in the natural environment, α1=[α UV1 ,α T1 ,α RH1 ] T Used to describe the upper and lower limits of ultraviolet intensity, temperature and relative humidity in the natural environment. Used to determine the time when the ultraviolet intensity, temperature and relative humidity in the natural environment reach the extreme value during periodic changes, ε=[ε UV ,ε T ,ε RH ] T Used to describe the random fluctuations of UV intensity, temperature and relative humidity in the natural environment, obeying the multivariate normal distribution MN(μ,Σ): μ=[0,0,0] T ; Based on the collected natural environment data, the maximum likelihood estimation method is used to estimate the above parameters.

5. The epoxy coating degradation prediction method considering the shielding effect of multiple environmental comprehensive influences according to claim 4 is characterized in that: The basic idea of the maximum likelihood estimation method is to maximize the likelihood function of the data observed under these parameters by selecting a set of parameters. The likelihood function is expressed as the product of the probabilities of each sample, which is: Its goal is to find the parameter θ so that the likelihood function L(θ) reaches its maximum value.

6. The method for predicting epoxy coating degradation under the shielding effect of multiple environmental comprehensive influences according to claim 5 is characterized in that: After obtaining the parameter estimation value through the maximum likelihood estimation method in step 4, the Monte Carlo simulation method is used to predict the degradation state of the epoxy coating under the field service environment. Specifically, a large number of multivariate normally distributed random numbers are randomly generated according to the results of the parameter estimation, and a large number of natural environment simulation data sequences are generated. The natural environment simulation data sequences are input into the trained random forest model based on particle swarm algorithm optimization to output the degradation state of the epoxy coating in the natural environment; the degradation interval is determined according to the quantile of the degradation amount at different degradation moments, and the interval prediction result of the degradation state of the epoxy coating is obtained. According to the 50% quantile, the point estimate of the mean value of the degradation state at different times is obtained, and finally the degradation state of the epoxy coating under the field service environment is predicted.

Images