Correlation Study On The Morphology Characterization Of Corroded Steel Bars And Their Mechanical Properties Based On 3D Scanning Technology

Reinforced concrete structures are widely used in infrastructure.Due to the diversity of service environments and the unevenness of pores in concrete structures,it is easy to cause uneven corrosion of internal steel bars,which reduces the service life of building structures and results in huge economic property losses.The statistical study of the characteristics of the appearance of corroded steel bars and their effect on the deterioration of the mechanical properties of corroded steel bars are beneficial to our assessment on the durability of concrete structures.The main work and research results of this paper are given as follows:Firstly,in this paper the steel bars in the corroded concrete are used as samples,which are semi-immersed and energized to accelerate corrosion.Based on the three-dimensional scanning technology and MATLAB platform,the average（volume）corrosion rate,critical corrosion rate,maximum pit depth,uneven coefficient R（ratio of average cross-sectional area to critical cross-sectional area）and other corrosion parameters are obtained.Through statistical regression analysis,the relationship between the above corrosion parameters and the quality corrosion rate are established.By using probabilistic statistics to analyze the obtained test data,this study finds that the maximum pit depth follows the extreme I-type distribution,the ratio of the maximum pit depth to the average pit depth follows the lognormal distribution,and the uneven coefficient R follows the extreme value Type II distribution.Based on fractal theory and statistics,two parameters characterizing the degree of unevenness of corroded steel bars are obtained:characteristic profile parameters and overall unevenness coefficient.In addition,the static tensile and dynamic characteristics tests are carried out on the corroded steel bars.The multiple indicators of average corrosion rate,critical corrosion rate,cross-section difference rate（critical corrosion rate-average corrosion rate）,uneven coefficient,characteristic profile parameters and overall unevenness coefficient are used to evaluate the influence of the mechanical properties of corroded steel bars.Through regression analysis,it is found that both the average corrosion rate and the critical corrosion rate can well evaluate the yield load or limit load of the corroded steel bar,which decreases linearly with the increase of the corrosion rate.The critical corrosion rate is used to evaluate the actual strength of the corroded steel bar,it shows that corrosion does not change the actual yield strength or limit strength of the steel bar.Nominal yield strength,nominal limit strength and ductility show a decreasing trend with the increase of various corrosion parameters.Among them,the nominal yield strength and nominal limit strength exhibit a strong linear degradation with the increase of unevenness coefficient and cross-section difference rate.With the increase of unevenness coefficient and cross-section difference rate,the ductility shows a strong exponential degradation.Thus,the unevenness coefficient and cross-section difference rate show a better result compared to other parameters of rust characteristic.Through dynamic characteristics tests,it shows that the natural frequency of corroded steel bars decreases linearly with the increase of the mass corrosion rate.The higher the modal order,the smaller the reduction rate.Moreover,the solid model created by 3D scanning is imported into finite element analysis software for uniaxial tensile simulation,which can better restores the mechanical properties of corroded steel bars and shows the model reliability.The stress distribution of the corroded steel bar is studied,and the nodal stress of the corroded steel bar is extracted.The result shows that the minimum cross-sectional area of corroded rebar is not necessarily the location of the maximum stress of corroded rebar,and the maximum stress of corroded rebar is determined by the minimum cross-sectional area and the shape of the rust pit.In this study,the random forest algorithm,a machine-learning based method,is used to predict the mechanical properties of corroded steel bars.90%of the 120 data sets of corroded steel bars are used as the training sets,and 10%of 120 data sets are used as the test sets.The characteristic parameters of the corroded steel bars are used as the inputs,and the mechanical properties of the corroded steel bars are used as the outputs.The result shows that the prediction results are the most accurate when all the corrosion characteristic parameters are taken into account（average corrosion rate,critical corrosion rate,uneven coefficient,overall uneven coefficient,maximum pit depth,length of the corrosion section）.By removing the weakest correlation factor of length of the corrosion section and combining randomly the rest 5 corrosion characteristic parameters,it is shown that we can obtain the satisfactory prediction results when more than or equal to 4 corrosion characteristic parameters are used as inputs.As of the case 3 corrosion characteristic parameters are used as inputs,the combination of R+Ra+₈（6）can get best prediction results.When two corrosion characteristic parameters are used as inputs,the combination of R+Ra or_（60）+R can get best prediction results.Besides,by comparing the prediction results of the random forest algorithm with the support vector machine,it is found that the prediction results of the random forest algorithm are significantly better than the support vector machine.