Abstract
The practical and accurate method for corrosion life prediction and anti-corrosion design of automobiles have been newly developed. A number of case studies on corrosion occurrence from automobiles aged from 4 to 10years in a real environment were collected. The neural-network-based nonlinear multivariate analysis method was applied to derive implicit relations of corrosion in automobiles from the collected case studies. Firstly, the method was used for the corrosion occurrence prediction about the SILL OUTER parts. Success rate of the prediction was over 90%. Secondly, the method was used to predict the E-coat thickness on the inner side of the SILL OUTER parts required for corrosion protection. Actual and estimated E-coat thickness values agreed reasonably well with each other, i.e. with a high linear correlation factor of 0.9.
