Research On Big Data Of Atmospheric Corrosion Of Carbon Steels And Action Law Of Main Influencing Factors

Due to the diversity of atmospheric environments,the distribution and impact of the main environmental factors which affect atmospheric corrosion(i.e.temperature,humidity,irradiance,chloride ions,salinity,pollutants)are significantly different in different macro and micro environments,and the strength of these impacts in different time periods are not the same.The diversity of environments and the dynamically changing laws of corrosion behaviors should be considered to understand the atmospheric corrosion of carbon steel.Based on the dynamic understanding of atmospheric corrosion behaviors,a new big data system of atmospheric corrosion was designed in this study,which achieved minutely and continuous monitoring of corrosion current of carbon steels,temperature and relative humidity in outdoor environments.This system is able to collect 500,000 pieces of data per year at any monitoring location,and this system was utilized to study the atmospheric corrosion of carbon steels.Based on the data collected,the main environmental factors which affected atmospheric corrosion were explored,and the action law of the rust layers on critical relative humidity of atmospheric corrosion in outdoor environments was revealed.Meanwhile,the atmospheric corrosion data sets with dynamic characteristics was used to predict atmospheric corrosion of carbon steel by different data mining methods,and the best corrosion prediction model was selected.In this study,the expansion of the covered corrosion products and effect of the galvanic acceleration on the corrosion sensor surface hardly made the sensor a substantial impact.The corrosion state of the carbon steels on the sensor could be considered basically the same as that of the corrosion coupons during the test.A transformational formula was existed between the output current of the corrosion sensor,the output electric quantity of the corrosion sensor and the corrosion rate of the corrosion coupons.The formula was suitable for a rapid assessment of the resistance to atmospheric corrosion of carbon steels and weathering steels in and above C2 level,which proved that the corrosion monitoring technology has the potential to replace traditional corrosion coupons on the study of atmospheric corrosion occurred on the low alloy steels.The environmental corrosivity in outdoor environments was highly dynamic.In the 1-year monitoring period,the environmental corrosivity was distributed from C1 to CX.During the initial atmospheric corrosion of the carbon steels,the most impactful climatic factor was rainfall except for an arid environment.The effect of relative humidity for the initial atmospheric corrosion was stronger than that of temperature,and the atmospheric corrosion usually occurred at night.The influence of the temperature on atmospheric corrosion was less reflected in the promotion effect of the accelerated reaction by high temperature.Temperature and the pollutants in urban atmospheric environment were considered as the secondary factors.The growth of the rust layer on the corrosion sensor was the most important corrosive factor affecting the output current value of the corrosion monitoring sensor.The growth of rust layer promoted the atmospheric corrosion caused by relative humidity.After a long-term testing of atmospheric corrosion of carbon steel,the top five corrosive factors influencing atmospheric corrosion in marine environment were the rust layer,wind speed,rainfall rate,relative humidity and the chloride ions.With the prolongation of the exposure time,the importance of wind speed gradually decreased,and the role of rust layer and chloride became more and more prominent.In the analysis of the multi-dimensional corrosion big data of carbon steels,the random forest model was able to correctly analyze the quantitative effects of various corrosion factors on carbon steel,and predict the atmospheric corrosion.The predicting accuracy of the random forest model was obviously better than the back propagation neural network and support vector regression models in predicting the atomospheric corrosion of carbon steels under outdoor environments.On the basis of the random forest algorithm,a new iterative random forest model was established considering the growth of rust layer.As a result,the accuracy of prediction was improved.The critical relative humidity of atmospheric corrosion of carbon steel refers to the relative humidity when atmospheric corrosion occurs.The thickness and phase structure of the rust layers formed in the outdoor environments showed obvious influence on it.If the structure of the rust layer was unchanged,the critical relative humidity would decrease with the increase of the thickness in the statistical range,which promotes the atmospheric corrosion caused by the change of temperature and humidity.The rust layer structure with γ-FeOOH as the main phase in the outer,middle and inner layers showed the strongest promoting effect among the four statistical rust layer structures.The rust layer structure with γ-FeOOH as the main phase while with α-FeOOH as the main phase in the middle and inner layers had the weakest promoting effect.Finally,The key effect of the changed phase structure on the improvement of the corrosion resistance for weathering steels was analyzed.The increased content of α-FeOOH in the rust layer effectively blocked the chloride ions from the outer rust layers and prevented a decrease in critical relative humidity for corrosion.Compared with the traditional method of corrosion coupons,the big data system of atmospheric corrosion could effectively shorten the evaluation period of corrosion resistant materials.