*4.5. Quality Evaluation*

For the correct functioning of the model, data were normalised. According to the previous criteria, the most similar options are shown. The best way to show the results of this last model is using an application example, which is presented in Table 7. The quality row shows the percentage assessing the prediction's quality. The first column represents all input values of the example. The next three columns show the most similar real results in the database.


**Table 7.** Results of the example case, using the distance model.

Results obtained above show high prediction reliability. Cases similar to the one under study have been found in the database. The model could also give a satisfactory result for a case that is not included in the database. Ideally, the results obtained with the proposed methodology should be compared with the results obtained with existing methods in the literature. However, since the innovative premise of this study is based on adapting the input variables to avoid the need for pollutant-specific data, such a comparison cannot be made. One of the differentiating factors of this classifier model is that to obtain a corrosion loss rate, values for pollutant concentrations are not needed. Consequently, it may be concluded that the different algorithms developed are a good alternative for technicians and engineers to make informed decisions based on their level of risk acceptance. To sum up, given a specific location and based on the available data, these models can determine the Zn-coating thickness needed for a successful short- and long-term corrosion resistance, providing the most probable, optimistic, and pessimistic predictions.

#### **5. Conclusions**

In the present work, various models for predicting galvanised coated steel corrosion damage of metal structures exposed to weathering have been developed. The following conclusions can be drawn from this research.

The application of a supersom algorithm is considered for first-year corrosion prediction, which allows categorising any environment while obtaining a predicted value, with satisfactory results. In the cases when no experimental data are available, the model can be an alternative to the conventional informative method based on pollutant input variables. The model presented in this work could help civil engineering companies to optimise the ratio between the minimum coating required and maximum service life, thus contributing to a significant lifetime extension of steel structures.

The main limitation of the model is that it lacks statistical metrics to evaluate the performance. To solve this and explore the performance and quality of the predictions, a quality model based on Euclidean distances was proposed. A long-term corrosion prediction was also optimised based on standards ISO 9224:2012 formula and the exponential coefficient with Newton's method.

To cover all different atmospheric environments, more specific characterisations are required. The future research will focus on including the development of physical variables, such as wind speed and wind direction. It is also important to feed the model with more

examples from the lesser-represented categories, as there are notable differences between C3/C4 categories and the remainder of the cases. Adding new metallic materials will also be explored, following the same methodology, possibly leading to the development of new prediction models.

**Author Contributions:** Conceptualisation, F.O.-F. and A.F.-I.; methodology, G.A.-I. and M.D.-P; validation, G.A.-I. and M.T.-C; writing—original draft preparation, M.T.-C.; writing—review and editing, M.T.-C. and M.D.-P.; supervision, F.O.-F. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Conflicts of Interest:** The authors declare no conflict of interest.
