Sistema automático de detección del estado de corrosión en aceros inoxidables austeníticos

Laburpena

The deterioration of the materials due to corrosion has a great impact on the economy, the public health and the environment. The corrosion prevention and control is highly complex since many factors are involved in this process. The electrochemical tests have been considered one of the most useful tools to be applied in corrosion studies. However, these techniques do not provide any method to analyse the corrosion behaviour of the material automatically, as a function of the environmental factors. In this Thesis, models based on artificial intelligence techniques are presented to identify, by an automatic way, the pitting corrosion status of AISI 316L austenitic stainless steel. Chloride ion concentration, pH and temperature have been the environmental factors considered in this work, since they are the most critical factors in this type of corrosion. The models have been developed based on the experimental data obtained from the European project called ¿Avoiding catastrophic corrosion failure of stainless steel¿ (RFSR-CT-2006-00022) ¿CORINOX. Multiple comparison tests analysis is proposed in order to determine the optimal configuration of each technique. The results, up to 98.1% sensitivity for artificial neural networks and 99.5% specificity for support vector machines, demonstrate the efficiency of the proposed models to be applied in pitting corrosion modelling of AISI 316L stainless steel.