Improved prediction of the corrosion behaviour of car body steel using a Kohonen self organising map

In a highly competitive market, an optimised processing plays a key role for the future of our industry. Intelligent processing is only feasible, when the materials and products are controlled directly during and after processing. Traditional methods for controlling corrosion are often very time consuming and destructive (e.g. corrosion under paint, delamination of paints on metals), Therefore these tests are restricted to a small number of samples and give only limited representative information of the investigated sample. In previous papers it has been shown, that diffuse reflectance spectroscopy can be used to describe such a complex behaviour like corrosion tendency of car body steel. This is an on line method, that is fast, contact free and gives statistical sufficient information. Due to the overlap of diffuse and specular information of the spectra, modern classification methods as neural networks must be applied. Additional difficulties appear, as the standard reference methods for corrosion evaluation like salt spray tests etc. give data only with a statistical variation. The authors look at the benefits of classification of the spectra by Kohonen nets