CN113588528A - Metal atmospheric corrosion prediction method considering dynamic characteristics of natural environment - Google Patents

Abstract

The invention provides a metal atmospheric corrosion prediction method considering the dynamic characteristics of the natural environment, which is realized by the following steps: first: establishing an environmental impact law model of atmospheric corrosion; second: estimating and selecting parameters of the environmental impact law model of atmospheric corrosion; three: Establish an atmospheric natural environment model considering dynamic, correlation and random characteristics; 4: Estimation of environmental model parameters and model verification; 5: Establish a corrosion prediction model under the environmental conditions of use; Through the above steps, it can be seen that the present invention comprehensively considers the atmosphere The corrosion influence law of multiple environmental factors in the natural environment can be combined with the information criterion to establish the optimal corrosion influence law model. The factors are more comprehensive and the model is more accurate; the natural environment information is more abundant, and the modeling results are more accurate. The actual use environment is closer, the method is scientific, and the craftsmanship is good. It can be applied to different environmental conditions and has high theoretical and application value.

Description

A metal atmospheric corrosion prediction method considering dynamic characteristics of natural environment

technical field

The invention provides a metal atmospheric corrosion prediction method considering the dynamic characteristics of the natural environment, which relates to an atmospheric corrosion prediction method considering the dynamic, correlation and random characteristics of the natural environment. The method mainly includes an environmental impact law model and a natural environment. There are two parts of the model. Among them, the environmental influence law model is a quantitative relationship model between environmental influence factors (such as temperature, relative humidity, pollutants, etc.) and corrosion performance parameters (such as corrosion weight loss, corrosion rate, etc.), and the natural environment model is used for Describe the environmental influences of dynamically correlated stochastic changes. By fusing the above two models, the atmospheric corrosion prediction method under the actual use environment conditions can be established. This method is suitable for environmental corrosion classification and evaluation, anti-corrosion design and material selection, environmental analysis and prediction and other fields.

Background technique

Metal materials are the most widely used materials, and the economic losses caused by corrosion are very serious under the actual atmospheric natural environment conditions. Therefore, it is necessary to study the influencing factors and corrosion influence rules of metal materials in the atmospheric natural environment, explore the change law of the corrosion performance of metal materials, predict the change law of the corrosion performance of metal materials in different atmospheric natural environments, and evaluate the environmental corrosion and anti-corrosion design. Work is very important.

However, there are still two deficiencies in the related research work at home and abroad. On the one hand, the atmospheric natural environment model is unreasonable. For the quantitative description of atmospheric natural environment factors, the three commonly used modeling methods, namely the mean method, the distribution method and the time series method, have their own characteristics and limitations. The mean value method cannot completely describe all the information of the environment, nor can it reflect the nonlinear characteristics of environmental effects. The distribution method is not suitable for the situation with large dynamic changes of factors, and the time series method still needs to be improved for the modeling of related environmental factors. On the other hand, the model of the influence of corrosion in the atmospheric natural environment is not accurate. For different metal materials, the main environmental influence factors may be different, generally including temperature, humidity and pollutants. There may be complex interactive effects between these different factors. Therefore, it is necessary to determine the types of major environmental impact factors for different materials and different natural environment conditions. At the same time, for the description of environmental factors, the selection of environmental variables in most of the existing corrosion models is unreasonable, and the use of annual average values to describe the natural environment is too simplified, and for the model of corrosion influence laws, the existing models mostly use linear or generalized linear It cannot accurately describe the complex nonlinear influence laws of comprehensive environmental factors.

Based on this, the present invention studies and establishes a metal atmospheric corrosion performance prediction method including the atmospheric natural environment model and the atmospheric environment factor corrosion influence law model, which accurately describes the atmospheric natural environment conditions, comprehensively reflects the atmospheric corrosion influence law, and scientifically predicts corrosion. The performance change law has important theoretical value and engineering significance.

SUMMARY OF THE INVENTION

(1) purpose of the present invention :

The purpose of the present invention is to provide a metal atmospheric corrosion prediction method considering the dynamic characteristics of the natural environment, which aims at the atmospheric corrosion problem of metals, considers the dynamic, correlation and random characteristics of the natural environment, establishes an atmospheric natural environment model, and accurately Describe the atmospheric natural environment factors; at the same time, according to the main environmental factors affecting the metal atmospheric corrosion process, a comprehensive environmental factor corrosion influence law model is established to comprehensively reflect the corrosion influence law of the atmospheric natural environment, and then the natural environment model and the influence law model are integrated. A corrosion prediction model under dynamic atmospheric and natural environment conditions is formed to scientifically predict the change law of corrosion performance.

(2) Technical solution :

The present invention needs to establish the following basic settings:

Setting 1: The content of the present invention is mainly aimed at the atmospheric corrosion problem of metal materials. The main environmental influencing factors include atmospheric temperature, relative humidity, and pollutant concentration, and the influence of other environmental factors such as rainfall, irradiation, wind speed, etc. is not considered for the time being;

Setting 2: The present invention does not consider the interaction effect caused by the joint action of different environmental factors, so the influence law model of environmental factors does not include interaction terms;

Setting 3: Temperature, relative humidity, and the corrosion effects of pollutants (sulfur dioxide and chloride ions) are described by the Arrhenius model, the Peck model, and the dose-response model, respectively. For detailed model forms, see step one;

Setting 4: The comprehensive environmental factor model can be integrated through the models of various environmental factors, or established by generalized generalized linear, generalized log-linear, and generalized Eyring models. For the detailed model form, see step 1;

The above Arrhenius model is an empirical formula created by Arrhenius of Sweden in 1889 to describe the acceleration law of temperature; the Peck model follows the power function form and is used to describe humidity. The empirical formula of the acceleration law; the generalized Eyring model is the empirical formula of the acceleration law that comprehensively considers multiple environmental factors including temperature;

The method proposed in the present invention mainly includes three aspects of research content: atmospheric corrosion environmental impact law modeling and parameter estimation, atmospheric natural environment modeling and parameter estimation, and ambient atmospheric corrosion modeling and prediction. Based on the above basic settings, the present invention The proposed method for predicting metal atmospheric corrosion considering the dynamic characteristics of the natural environment is characterized in that: it is realized through the following steps:

Step 1: Establish the environmental impact law model of atmospheric corrosion

First, determine the main influencing factors based on the actual use environment of the metal material. Next, establish the corrosion influence law model for the selected environmental factors respectively. Influence law model;

The specific steps are:

I. Identifying key environmental impact factors

Generally speaking, in addition to temperature and relative humidity, it is also necessary to determine whether the influence of atmospheric pollutants needs to be considered according to the type of atmosphere; if it is an industrial atmospheric environment, the influence of sulfur dioxide needs to be considered, and if it is a marine atmospheric environment, it is necessary to consider the chloride ion If it has the characteristics of industrial atmosphere and marine atmosphere at the same time, the joint influence of sulfur dioxide and chloride ions should be considered at the same time. If it is a rural atmospheric environment, the influence of air pollutants will not be considered, and the specific selection can be made according to the following;

II. Establish a model of the influence laws of various environmental factors

For relative humidity, the influence law model is the Peck model:

r(RH)=a·RH ^b (1)

Among them, r(RH) is the corrosion rate, RH is the relative humidity (%) of the environment, and a and b are constants;

For temperature, the influence law model is the Arrhenius model:

Among them, r(T) is the corrosion rate, T is the ambient temperature (K), d is a constant, e=E _a /K, E _a is the activation energy of the reaction, which can be estimated from the test data, and K is the Boltzmann constant;

For sulfur dioxide concentration, the model of its influence law is a dose-response model:

r(S)=(1+f·S) ^g (3)

where r(S) is the corrosion rate, S is the sulfur dioxide concentration (μg.m ^-3 ), and f and g are constants;

For the concentration of chloride ion deposition rate, the influence law model is also the dose-response model:

r(Cl)=(1+h·Cl) ^k (4)

where r(Cl) is the corrosion rate, Cl is the chloride ion deposition rate (mg.m ^-2 .day ^-1 ), and h and k are constants;

III. Establish a comprehensive environmental impact law model

Multiply the influence law models of the above environmental factors to establish a comprehensive environmental influence law model, the form of which is as follows:

r(RH,T,S,Cl)=r(RH)·r(T)·r(S)·r(Cl) (5)

Formula (5) is the basic form for calculating the comprehensive influence of various environmental factors on atmospheric corrosion, and it can be extended to include the influence of more environmental factors, such as wind speed or solar radiation; Under certain conditions, the accuracy of the model established by formula (5) cannot be guaranteed. Therefore, the following two models can also be used:

Generalized log-linear model:

r(RH,T,S,Cl)=a·exp(b·RH+e·T+f·SO ₂ +h·Cl) (6)

Generalized Eyring model:

Equations (5)-(7) are general models. When the influence of other environmental factors needs to be considered, the same method can be used to include their influence laws in the model;

Step 2: Estimation and selection of model parameters for the environmental impact law of atmospheric corrosion

After the comprehensive environmental model is established, the parameters of the model need to be estimated to determine its quantitative influence law. At the same time, the optimal model needs to be selected from the alternative models. The selection is based on the Akaike Information Criterion (AIC). , the amount of model information determined by the corrected Akaike Information Criterion (AIC _C ) and Bayesian Information Criterion (BIC);

Among them, AIC, AIC _C , and BIC are all standards for measuring the goodness of statistical model fitting; AIC criterion was founded and developed by Japanese statistician Hiroji Akaike, so it is also called Akaike Information Content Criterion; AIC _C criterion is the basis of AIC Improvements have been made on the above, which can effectively avoid the estimation error in the case of small samples; the BIC criterion is called the Bayesian information criterion, and compared with AIC, it can effectively avoid the problem of overfitting in the case of large samples;

The specific steps are:

I. Model parameter estimation

The model form of equations (5)-(7) is relatively simple, and the parameter estimation method can be carried out by generalized linear regression or maximum likelihood method; this step can be directly realized by MATLAB software, which is very simple and will not be expanded here; (wherein , the "MATLAB software" is a commercial mathematical software produced by the American MathWorks company for data analysis);

II. Model Selection

Considering the model fitting accuracy and model complexity comprehensively, the optimal model is determined by the amount of information; the amount of information that can be used includes AIC, AIC _C and BIC, and the specific calculation formula is:

Among them, M is the number of parameters in the model, N is the sample size when fitting parameters, that is, the number of test data, RSS is the residual sum of squares, S _I and S _I ' are the first (I=1, 2,...,N) experimental measurements and model predictions;

AIC, AIC _C and BIC are a class of statistics that comprehensively consider model complexity and model prediction accuracy. When selecting models, the model with the smallest AIC, AIC _C or BIC value is used as the optimal model;

Step 3: Establish an atmospheric natural environment model considering dynamic, correlation and random characteristics

Taking temperature, humidity, and sulfur dioxide concentration as examples, in the actual natural environment, these environmental factors all change dynamically, and have periodic and random fluctuation characteristics; in order to accurately describe this dynamic change that includes both periodicity and randomness characteristics, establish the following time-varying function model

用于描述的季节性变化特征；A₁＝[A_T1,A_RH1,A_S1]为年波动幅值，τ₁＝[τ_T1,τ_RH1,τ_S1]和

为波动周期和相位；

用于描述日变化情况；ε(t)＝[ε_T(t),ε_RH(t),ε_S(t)]为随机波动项，服从特定的统计分布；通常情况下，ε(t)服从均值为0，方差为σ²(t)的正态分布，即ε(t)～N(0,σ²(t))；Among them, ξ(t)=(T(t), RH(t), S(t)) represents the ambient temperature, relative humidity and sulfur dioxide concentration at time t, and A ₀ =[A _T0 ,A _RH0 ,A _S0 ] is constant,

Seasonal variation characteristics used for description; A ₁ =[A _T1 ,A _RH1 ,A _S1 ] is the annual fluctuation amplitude, τ ₁ =[τ _T1 ,τ _RH1 ,τ _S1 ] and

are the fluctuation period and phase;

Used to describe the diurnal variation; ε(t)=[ε _T (t),ε _RH (t),ε _S (t)] is a random fluctuation term that obeys a specific statistical distribution; usually, ε(t) Obey the normal distribution with mean 0 and variance σ ² (t), namely ε(t)～N(0,σ ² (t));

Since the daily fluctuation term and the annual fluctuation term are periodic trigonometric functions with a mean value of 0, the mean value of each environmental factor is

Through the analysis of a large number of environmental data in different cities, it is found that the random fluctuation term of temperature and humidity can be accurately described by normal distribution, while the random fluctuation term of sulfur dioxide concentration can be described by lognormal distribution, which can be expressed as

和

分别表示温度和相对湿度在t时刻随机波动项的分布方差，θ_S(t)，

η_S(t)是t时刻二氧化硫浓度对数正态分布的参数；in,

and

respectively represent the distribution variance of the random fluctuation term of temperature and relative humidity at time t, θ _S (t),

η _S (t) is the parameter of the lognormal distribution of sulfur dioxide concentration at time t;

对于二氧化硫For temperature and relative humidity,

for sulfur dioxide

The analysis found that the stochastic fluctuation term of sulfur dioxide concentration in some locations can be described by an exponential distribution, that is,

ε _S (t)～E(λ _S (t),ζ _S (t)) (16)

λ _S (t) and ζ _S (t) are the rate parameter and location parameter of the exponential distribution of sulfur dioxide concentration at time t, respectively, then

Step 4: Environmental model parameter estimation and model validation

In step 3, the random term ε(t) of different environmental factors conforms to different statistical distributions; first, according to the different characteristics of ε(t), the parameters of the environmental model are estimated, and then the model data is generated by the fitting model, Compare with the original observation data to verify the accuracy of the model; the specific steps are as follows:

I. Environmental Model Parameter Estimation

ε(t)符合正态分布，公式(1)中的其他参数可用以下几步进行计算：For temperature and relative humidity, the seasonal variation of temperature is closely related to the revolution of the earth, while the diurnal variation depends on the rotation of the earth, so τ ₁ =1, τ ₂ =1/365; while for a large number of temperature observation data The analysis found that the annual minimum temperature generally occurs in mid-January every year, and the daily minimum temperature generally occurs around 2:00 in the morning. Therefore,

ε(t) conforms to a normal distribution, and other parameters in formula (1) can be calculated by the following steps:

则

i) Calculate the annual mean temperature

but

ii) Divide all the temperature data into twelve groups on a monthly basis, and calculate the monthly average temperatures T ₁ , T ₂ ,..., T ₁₂ , then

iii) Calculate the daily average temperature difference between day and night T′ in a year, then A _T2 =T′;

iv) Bring the parameter values calculated in the above three steps into the model to generate simulated data without random items that correspond to the original data recording time one-to-one. After subtracting the simulated data from the original data, the remaining data are still Each month is divided into twelve groups to obtain ε _T (t), and the variance of each group is calculated

与时间t的关系；v) Fit variance

relationship with time t;

For sulfur dioxide and chloride deposition rates, the environmental model parameters can be estimated through the following steps:

同时，每一天的二氧化硫浓度变化有两个高峰两个低谷，其两个高峰大约出现每日早9:00和晚21:00左右，这与每日城市中早晚高峰的时间基本吻合，期间机动车排放二氧化硫量较多，而在其他时间相对较少，因此，τ_z2＝0.5，

去除二氧化硫浓度数据中的年变化和日变化部分，发现其随机变化部分即ε(t)符合指数分布，且当二氧化硫浓度的日均值越大，日变化幅值和随机变化部分的波动也越大，因此日变化项A_z2(t)和ε_z(t)均随时间变化，假设A_z2(t)和ε_z(t)均和日均值成正比，则公式(12)中的其他参数可用蒙特卡罗方法进行估计，具体步骤为：In winter, due to heating and other reasons, the amount of fossil fuels burned is huge, and the emission of sulfur dioxide also increases. In summer, the amount of fossil fuels burned is less, and the amount of sulfur dioxide emissions decreases. The analysis found that the highest concentration of sulfur dioxide in a year occurs around early February, so τ _z1 = 365,

At the same time, there are two peaks and two troughs in the change of sulfur dioxide concentration every day. The two peaks appear at about 9:00 in the morning and 21:00 in the evening, which is basically consistent with the daily morning and evening peaks in the city. The amount of sulfur dioxide emitted by the motor vehicle is relatively large, while it is relatively small at other times. Therefore, τ _z2 = 0.5,

Removing the annual and daily variation of the SO2 concentration data, it is found that the random variation, ε(t), conforms to the exponential distribution, and the greater the daily mean value of SO2 concentration, the greater the fluctuation of the daily variation amplitude and the random variation. , so the diurnal variation terms A _z2 (t) and ε _z (t) both vary with time. Assuming that both A _z2 (t) and ε _z (t) are proportional to the daily mean, other parameters in formula (12) can be used Monte Carlo method to estimate, the specific steps are:

i) Calculate the annual mean value of sulfur dioxide and record it as

ii) the maximum value and minimum value of sulfur dioxide are recorded as z _min and z _max respectively;

的随机数

并将该值赋予A_z0作为蒙特卡罗方法计算的初始值，然后令

iii) generate a uniform distribution

random number of

and assign this value to A _z0 as the initial value calculated by the Monte Carlo method, then let

iv) Bring the parameters A _z1 , A _z2 (t) and λ _z (t) obtained in the previous step into the time-varying function model to generate N simulated values of sulfur dioxide concentration data, where N is equal to the modeled sulfur dioxide concentration data The total number of , count the number of these data falling into the k intervals [z _min ,z ₁ ],…,(z _i-1 ,z _i ],…,(z _k-1 ,z _max ], respectively, and calculate the statistic χ ²

Among them, n _i and n _i ′ are the numbers of simulated data and actual observed data falling into the i-th interval (z _i-1 , z _i ] respectively;

即为A_z0的估计值，将A_z0带入第三步的公式中，可得到A_z1，A_z2(t)，λ_z(t)的估计值；v) Repeat step 3 to step 4 m times, until the χ ² value obtains the minimum value that meets the accuracy requirement, and the corresponding

It is the estimated value of A _z0 . Bring A _z0 into the formula of the third step to obtain the estimated value of A _z1 , A _z2 (t), and λ _z (t);

II. Environment Model Validation

Specific steps are as follows:

i) Bring the parameter estimation results into the model, and simulate the environmental data for one year at three-hour intervals;

ii) Fit the probability density function (PDF) of the simulated data and the original data, and denote them as p _o and p _s , respectively;

iii) Calculate the model fitting error η by the following formula:

Among them, ξ _L and ξ _U are the lower and upper limits of the distribution of environmental factors ξ, respectively;

iv) If η<20%, the model passes the verification, if >20%, the method is not applicable;

Step 5: Establish a corrosion prediction model under ambient environmental conditions

Using the environmental impact law model and natural environment model obtained in steps 1 to 4, the following corrosion prediction model can be established:

Among them, C(τ) is the corrosion weight loss to time τ, r( ) is the corrosion influence law model, ξ(t)=[T(t), RH(t), S(t), Cl(t)] is the value of the environmental factor at time t, θ=[a,b,…,h] is the model parameter of the corrosion influence law;

In order to verify the accuracy of the prediction results of the model, the original observation data of environmental factors are used to calculate the true value of corrosion weight loss C _A (τ), and the calculation formula is as follows:

Among them, i is the order of observations in chronological order, and n _τ is the total number of observation data from 0 to τ;

Furthermore, the relative error γ between the predicted value and the actual value is calculated as the basis for judging the prediction accuracy of the model. The formula is:

Through the above steps, it can be seen that the present invention comprehensively considers the corrosion influence law of multiple environmental factors in the atmospheric natural environment, and can establish an optimal corrosion influence law model in combination with the information criterion, the factors are more comprehensive, and the model is more accurate; The natural environment model fully considers the actual characteristics of environmental dynamics, correlation and randomness. The natural environment information is more abundant, and the modeling results are more accurate. The established corrosion prediction method is closer to the actual use environment. The method is scientific and the process is good. It can be applied to different environmental conditions and has high theoretical and practical value.

(3) Advantages and efficacy :

The present invention is a metal atmospheric corrosion prediction method considering the dynamic characteristics of the natural environment, that is, a metal atmospheric corrosion prediction method considering the dynamic, relevant and random characteristics of the natural environment, and has the following advantages:

① The present invention comprehensively considers the corrosion influence law of multiple environmental factors in the atmospheric natural environment, and can establish an optimal corrosion influence law model in combination with the information criterion, the factors are more comprehensive, and the model is more accurate;

② The atmospheric natural environment model established by the present invention fully considers the actual characteristics of environmental dynamics, correlation and randomness, and the natural environment information is more abundant, and the modeling results are more accurate;

③ The corrosion prediction method established by the present invention is closer to the actual use environment, the method is scientific, and the manufacturability is good.

And can be applied to different environmental conditions, with high theoretical and practical value;

Description of drawings

Fig. 1 is a framework diagram of the method of the present invention.

Fig. 2 The comparison between the simulation results of the natural environment model and the original observations.

The prediction curve of corrosion weight loss in the case of Fig. 3.

Detailed ways

The present invention will be described in further detail below with reference to examples.

An accelerated corrosion test was carried out on a zinc alloy material under laboratory conditions, and the corrosion current I _corr was measured at different temperatures T and relative humidity RH. The test data are shown in Table 1.

Table 1 Materials Accelerated Corrosion Test Data

At the same time, taking Beijing's environmental data from 2007 to 2016 as an example, the actual effect of the atmospheric natural environment model included in the present invention is shown. Among them, the temperature and relative humidity data are from NOAA, the unit is °C and %, respectively, and a data is measured and recorded every 3 hours, and the sulfur dioxide data is from the China National Environmental Monitoring Center (CNEMC), the unit is μg.m ^-3 , every 1 Measure and record one data hourly. It should be noted that, in order to ensure the accuracy of the estimation of the parameters of the environmental model, if the original data allows, there should be no less than 6 data points per day, that is, at least one data is measured and recorded every 4 hours.

The technical framework of the corrosion prediction method involved in the present invention is shown in Figure 1, and the detailed steps are as follows:

Step 1: Establish the environmental impact law model of atmospheric corrosion

I. Identifying key environmental impact factors

This case involves only two environmental factors, temperature and relative humidity, which can be applied to corrosion prediction of metals in rural atmospheric conditions. Due to the limitation of experimental data, the influence of other environmental factors is not considered for the time being. In theory, this method can cover all the main influencing factors when there is sufficient experimental data to support it.

II. Establish a model of the influence laws of various environmental factors

For relative humidity and temperature, the Peck model and Arrhenius model represented by formulas (1) and (2) are used to quantitatively describe their influence laws;

III. Establish a comprehensive environmental impact law model

Based on the influence law model of each environmental factor in the previous step, the following comprehensive environmental influence law models are established respectively:

Combined model:

Generalized log-linear model:

r(RH,T)=a·exp(b·RH+e·T) (22)

Generalized Eyring model:

Step 2: Estimation and selection of model parameters for the environmental impact law of atmospheric corrosion

I. Model parameter estimation

The generalized linear regression method is used to estimate the model parameters in formulas (21)-(23), and the results are shown in Table 2;

Table 2 Parameter estimation results of the comprehensive environmental impact law model and model information

Model lna b e AIC AIC_C BIC Combination model -19.1 6.23 6140.7 -4.86 -3.66 -1.33 Generalized log-linear model -42.7 0.129 0.068 -39.55 -38.35 -36.02 Generalized Eyring Model -2.23 0.129 5993.4 -39.94 -38.74 -36.41

II. Model Selection

The information content value of each model is calculated by formulas (8)-(11), and the results are shown in Table 2. According to the information criterion, in this case, the AIC, AIC _C and BIC values of the generalized Eyring model are all the smallest, so it is the optimal model in this case;

Step 3: Establish an atmospheric natural environment model considering dynamic, correlation and random characteristics

In this case, only two environmental factors, temperature and relative humidity, are considered. Therefore, the natural environment model is:

Among them, ξ(t)=(T(t), RH(t)) represents the ambient temperature and relative humidity at time t;

Step 4: Environmental model parameter estimation and model validation

I. Environmental Model Parameter Estimation

Taking the natural environment of Beijing as an example, the estimation results of the parameters of the environmental model are shown in Table 3;

Table 3 Parameter estimation results of natural environment model and relative errors of the model

II. Environment Model Validation

The parameter estimation results in the above table are brought into the model to generate simulated environmental data, the data interval is one every 3 hours, a total of one year of data is generated, and the probability density function is fitted to all the generated data. The relative error value η between the simulated environmental data and the original environmental data is calculated by formula (19), and the results are shown in Table 3; it can be seen that for temperature and relative humidity, η is 8.7% and 5.2%, respectively, which proves that this The natural environment modeling method proposed by the invention has high modeling accuracy;

Step 5: Establish a corrosion prediction model under environmental conditions

Based on the modeling and calculation results of the above steps, the estimated parameters are substituted into the model, and the corrosion prediction model under the trial environment conditions is obtained as follows

in

Take τ=365, then C(τ) is the cumulative corrosion amount in one year, and at the same time, use the following formula to calculate the true value of the corrosion amount

Among them, the environmental data is observed and recorded every 3 hours, there are 8 times a day, and a total of 2820 times a year, so n _τ =2920.

The corrosion calculation results are shown in Figure 3; the predicted value of the model is the true value, and the relative error of the model prediction is xx%, which proves that the model has high prediction accuracy.

Claims (1)

1. A metal atmospheric corrosion prediction method considering the dynamic characteristics of the natural environment is provided with the following steps:

setting 1: aiming at the atmospheric corrosion problem of metal materials, environmental influence factors comprise atmospheric temperature, relative humidity and pollutant concentration, and other environmental factors are not considered;

setting 2: interaction effects generated by the combined action of different environmental factors are not considered, so that an influence rule model of the environmental factors does not contain interaction items;

setting 3: the temperature, the relative humidity and the corrosion influence rule of pollutants are respectively described by an Arrhenius model, a Peck model and a dose-response model, and the detailed model form is shown in step one;

setting 4: the comprehensive environment factor model is synthesized through models of all environment factors, or is established by adopting a general generalized linear, generalized logarithmic linear and generalized alling Eying model, and the detailed model form is shown in step one;

the method is characterized in that: the method is realized by the following steps:

the method comprises the following steps: establishing an environmental influence rule model of atmospheric corrosion

Firstly, determining influence factors based on the actual use environment of the metal material, then respectively establishing a corrosion influence rule model for the selected environment factors, and on the basis, comprehensively considering the combined action of all the factors and establishing a comprehensive environment factor influence rule model; the method comprises the following specific steps:

I. determining key environmental impact factors

In addition to temperature and relative humidity, it is also necessary to determine whether the effect of atmospheric pollutants needs to be considered according to the type of atmosphere; if the environment is an industrial atmospheric environment, the influence of sulfur dioxide needs to be considered, if the environment is a marine atmospheric environment, the influence of chloride ions needs to be considered, if the environment has the characteristics of industrial atmosphere and marine atmosphere, the common influence of sulfur dioxide and chloride ions needs to be considered, if the environment is a rural atmospheric environment, the influence of atmospheric pollutants is not considered, and the selection is specifically performed according to the following steps;

II, establishing an influence rule model of each environmental factor

For relative humidity, the model of the influence law is the Peck model:

r(RH)＝a·RH^b (1)

wherein R (RH) is the corrosion rate, RH is the relative humidity (%) of the environment, and a and b are constants;

for temperature, the law model of its effect is the Arrhenius model:

where r (T) is the corrosion rate, T is the ambient temperature (K), d is a constant, E ═ E_a/K，E_aThe reaction activation energy is estimated from test data, and K is a Boltzmann constant;

for sulfur dioxide concentration, the influence rule model is a dose-response model:

r(S)＝(1+f·S)^g (3)

wherein r (S) is the corrosion rate, and S is the sulfur dioxide concentration (μ g.m)^-3) F and g are constants;

for chloride ion deposition rate concentration, the influence law model is also a dose-response model:

r(Cl)＝(1+h·Cl)^k (4)

wherein r (Cl) is the corrosion rate, and Cl is the deposition rate of chloride ion (mg.m)^-2.day^-1) H and k are constants;

III, establishing a comprehensive environment influence rule model

Multiplying the influence rule models of the environmental factors to establish a comprehensive environmental influence rule model, wherein the form of the comprehensive environmental influence rule model is as follows:

r(RH,T,S,Cl)＝r(RH)·r(T)·r(S)·r(Cl) (5)

formula (5) is a basic form for calculating the comprehensive influence of each environmental factor on atmospheric corrosion, and meanwhile, in consideration of specific conditions which may occur under different situations, the accuracy of establishing the model by using formula (5) can not be guaranteed, so that the following two models are adopted:

generalized log-linear model:

r(RH,T,S,Cl)＝a·exp(b·RH+e·T+f·SO₂+h·Cl) (6)

generalized Eying model:

step two: estimation and selection of environmental influence law model parameters of atmospheric corrosion

After the comprehensive environment model is established, the model parameters need to be estimated so as to determine the quantitative influence rule, and meanwhile, the optimal model needs to be selected from the alternative models according to the Chichi information criterion AIC and the corrected Chichi information criterion AIC_CThe model information quantity determined by the Bayesian information criterion BIC; the method comprises the following specific steps:

I. model parameter estimation

The parameter estimation modes of the formulas (5) to (7) are carried out by adopting a generalized linear regression or a maximum likelihood method; the step is directly realized by MATLAB software;

model selection

Comprehensively considering model fitting precision and model complexity, and judging an optimal model by adopting the information quantity; the information amount includes AIC, AIC_CAnd BIC, the specific calculation formula is as follows:

wherein M is the number of parameters in the model, N is the sample size when fitting the parameters, i.e. the number of test data, RSS is the sum of squares of the residuals, S_IAnd S'_IRespectively an I test measured value and a model predicted value; 1,2, …, N;

step three: establishing an atmospheric natural environment model considering dynamic, relevant and random characteristics

In an actual natural environment, environmental factors such as temperature, humidity and sulfur dioxide concentration are dynamically changed, and the device has a periodic fluctuation characteristic and a random fluctuation characteristic; in order to accurately describe the dynamic change characteristics including periodicity and randomness, the following time-varying function model is established

Where ξ (t) ═ t, (t), rh (t), s (t)) represents the ambient temperature, relative humidity and sulfur dioxide concentration at time t, a₀＝[A_T0,A_RH0,A_S0]Is a constant number of times, and is,

seasonal variation characteristics for description; a. the₁＝[A_T1,A_RH1,A_S1]For annual fluctuation amplitude, τ₁＝[τ_T1,τ_RH1,τ_S1]And

the period and phase of the wave;

for describing the daily change; epsilon (t) ([ epsilon ]_T(t),ε_RH(t),ε_S(t)]A random fluctuation item obeys a specific statistical distribution; ε (t) obeys a mean of 0 and a variance of σ²Normal distribution of (t), i.e. ε (t) to N (0, σ)²(t))；

Since the daily fluctuation term and the annual fluctuation term are periodic trigonometric functions having a mean value of 0, the mean value of the environmental factors is

The random fluctuation terms of the temperature and the humidity are accurately described by normal distribution, and the random fluctuation terms of the sulfur dioxide concentration are described by lognormal distribution and are specifically expressed as

Wherein,

and

representing the variance, theta, of the distribution of the random fluctuation terms of temperature and relative humidity, respectively, at time t_S(t)，

η_S(t) is a parameter of the lognormal distribution of the sulfur dioxide concentration at the time t;

as for the temperature and the relative humidity,

for sulfur dioxide

It has been found by analysis that the random fluctuation term of the sulfur dioxide concentration can be described by an exponential distribution, i.e.

ε_S(t)～E(λ_S(t),ζ_S(t)) (16)

λ_S(t) and ζ_S(t) is the rate parameter and the position parameter of the exponential distribution of the sulfur dioxide concentration at time t, respectively, then

Step four: environmental model parameter estimation and model verification

In step three, the statistical distribution of random terms epsilon (t) of different environmental factors is different; firstly, estimating environmental model parameters according to different characteristics of epsilon (t), further generating model data by using a fitting model, comparing the model data with original observation data, and verifying the accuracy of the model; the method comprises the following specific steps:

I. environmental model parameter estimation

With respect to temperature and relative humidity, seasonal changes in temperature are closely related to the earth's revolution, while diurnal changes depend on the earth's rotation, so τ₁＝1，τ₂1/365; analysis of a large amount of temperature observation data shows that the annual minimum temperature occurs in the middle of one month every year, and the daily minimum temperature occurs at about two early morning hours, so that,

ε (t) follows a normal distribution, and the other parameters in equation (1) are calculated in the following steps:

i) calculating the annual average temperature

Then

ii) dividing all temperature data into twelve groups on a monthly basis, and calculating the average temperature T of each month₁,T₂,…,T₁₂Then, then

iii) calculating the mean day-night temperature difference T' of the year, then A_T2＝T′；

iv) substituting the parameter values obtained by the three steps into a model to generate simulation data which is in one-to-one correspondence with the recording time of the original data and is not provided with random items, and dividing the rest data into twelve groups according to the standard of each month after subtracting the simulation data from the original data to obtain epsilon_T(t) calculating the variance of each group

v) variance of fit

The relation with time t;

for the deposition rates of sulfur dioxide and chloride ions, the environmental model parameters are estimated by the following steps:

the highest sulfur dioxide concentration occurs in the beginning of the second month in one year, so τ_z1＝365，

Meanwhile, the sulfur dioxide concentration change in each day has two peaks and two valleys, wherein the two peaks appear about 9:00 and 21:00 earlier and later every day, which is consistent with the peak time of the morning and evening in the city every day, during which the vehicle emits more sulfur dioxide, and relatively less sulfur dioxide in other times, therefore, tau_z2＝0.5，

Removing annual change and daily change parts in the sulfur dioxide concentration data, finding that the random change part, namely epsilon (t) accords with exponential distribution, and when the daily mean value of the sulfur dioxide concentration is larger, the daily change amplitude and the fluctuation of the random change part are larger, so the daily change item A is larger_z2(t) and ε_z(t) all vary with time, let A_z2(t) and ε_z(t) are all in direct proportion to the daily average, then other parameters in the formula (12) are estimated by a Monte Carlo method, and the specific steps are as follows:

i) calculating the annual average value of sulfur dioxide and recording the value

ii) the maximum and minimum values of sulfur dioxide are respectively denoted as z_minAnd z_max；

iii) generating a obedient uniform distribution

Random number of

And assigning this value to A_z0As initial values calculated by the Monte Carlo method, and then let

iv) subjecting the individual parameters A obtained in the preceding step to_z1，A_z2(t)，λ_z(t) substituting into the time-varying function model to generate N simulated values of sulfur dioxide concentration data, wherein N is equal to the total number of sulfur dioxide concentration data to be modeled, and counting the data to respectively fall into [ z [_min,z₁],…,(z_i-1,z_i],…,(z_k-1,z_max]The number of k intervals is calculated, and statistic chi is calculated²

Wherein n is_iAnd n'_iThe simulation data and the actual observation data respectively fall into the ith interval (z)_i-1,z_i]The number of (2);

v) repeating the steps three to four m times until the value of x²The value is the minimum value meeting the precision requirement, and the corresponding value at the moment

Is A_z0Is estimated as A_z0Substituting into the formula of the third step to obtain A_z1，A_z2(t)，λ_z(t) an estimate of;

verification of environmental model

The method comprises the following specific steps:

i) bringing the parameter estimation result into a model, and simulating and generating environmental data of one year at intervals of three hours;

ii) fitting the Probability Density Function (PDF) of the simulated data to the original data, and respectively recording as p_oAnd p_s；

iii) calculating the model fitting error η by the following formula:

wherein ξ_LAnd xi_URespectively the lower limit and the upper limit of the xi distribution of the environmental factor;

iv) if η < 20%, the model is validated, if > 20%, the method is not applicable;

step five: establishing a corrosion prediction model under the condition of use environment

And (3) establishing the following corrosion prediction model by utilizing the environmental influence rule model and the natural environment model obtained in the first step to the fourth step:

wherein C (tau) is corrosion weight loss until the moment tau, r (-) is a corrosion influence rule model,

xi (t) ═ t (t), rh (t), s (t), cl (t) ] is the value of the environmental factor at time t, and θ ═ a, b, …, h is the corrosion influence law model parameter;

in order to verify the accuracy of the model prediction result, the actual value C of the corrosion weightlessness is calculated by utilizing the original observation data of the environmental factors_A(τ), the calculation formula is as follows:

wherein i is the sequence of observed values according to time sequence, and n_τThe total number of observed data in time from 0 to tau;

further, a relative error γ between the predicted value and the true value is calculated as a basis for judging the prediction accuracy of the model, and the formula is as follows:

Images