*4.4. Long-Term Corrosion Prediction*

Once the first-year corrosion rate provided by the supersom model is known, the long-term loss can be identified thanks to the optimised Equation (1). Table 6 shows the different values obtained by this optimisation method for each of the corrosivity categories.



Figure 14 compares the distribution of relative errors of both models. The nonlinear regression relative error is represented by a solid black line and the standard formula's relative error (ISO 9224) by a blue dashed line. A more uniform distribution is achieved in the nonlinear regression model.

**Figure 14.** Comparison between Nonlinear regression and standard's formula relative errors.
