Reanalysis of Thickness Loss Data for Weathering Steel

The current ASTM Standard Guide G101 has a method for estimating the atmospheric corrosion resistance of low-alloy steels from the content of five elements: Cu, Ni, Cr, Si, and P. The ASTM method was developed by fitting a quadratic model to thickness loss data from specimens that were made from 270 different steel compositions and were exposed to the atmosphere for 15.5 years. The quadratic structure resulted in a model that provides an irrational measure of the effect of the most important alloying element. Additionally, the use of the correlation coefficient as the single criterion for selecting predictor variables led to inaccurate coefficients. The reanalysis of the writers, using multiple criteria for selecting the best predictor variables, yields more accurate coefficients and a level of explained variance of nearly 80%. While the multiple criteria analysis of the writers does not sacrifice prediction accuracy to achieve rationality, the problem inherent to the quadratic model cannot be overcome altogether in low-alloy steels containing more than about 0.40% copper. The corrosion index of Guide 101 should not be used to decide whether a low-alloy steel qualifies as a weathering steel.