Crevice corrosion: a neural network approach

Abstract

Crevice corrosion of stainless steels and related alloys in near neutral chloride containing environments is a very widespread form of localised corrosion in industrial plants where sea water is a common coolant medium and the complexity gives rise to many types of potentially dangerous crevices. Crevice corrosion, like other modes of localised corrosion, is a stochastic process which should be treated using probability laws and concepts. Many test methods have thus far been proposed and several different models have been developed to study crevice corrosion phenomena. However, much remains to be done and to be understood before this problem can be fully solved. The complexity of the processes occurring during the onset and development of crevice corrosion, as well as the poor reliability of the above mentioned test methods, have led to the development of artificial intelligence systems. Among these, the artificial neural network (NN) system appears to be a powerful tool to tackle situations in which a large number of data are available but no simple models are derived. In the present paper an attempt to develop a NN for rapid predictions of crevice corrosion performance of alloys in chloride containing environments is described.