作者: Neda Narimani , Bahram Zarei , Hesam Pouraliakbar , Gholamreza Khalaj
DOI: 10.1016/J.MEASUREMENT.2014.11.011
关键词:
摘要: Abstract Artificial neural networks with feed forward topology and back propagation algorithm were employed to predict the effects of chemical composition corrosion cell characteristics on both current density potential microalloyed pipeline steels. Doing this, compositions comprising “carbon” , “magnesium” “niobium” “titanium” “nitrogen” “molybdenum” “nickel” “aluminum” “copper” “chromium” “vanadium” “carbon equivalent” (all in weight percentage) along “reference electrode” “scan rate” “temperature ”, “relative pressure oxygen” “ purged CO 2 ” chloride ion as well as, bicarbonate concentration considered together input parameters network while “corrosion density” potential” outputs. For purpose constructing models, 87 different data gathered from literatures wherein examinations performed. Then randomly divided into training, testing validating sets. Scatter plots statistical criteria “absolute fraction variance (R )” “mean relative error (MRE)” used evaluate prediction performance universality developed models. Based analyses, proposed models could be further practical applications monitoring