Experiments, science and intuition in the development of models for the corrosion of steel infrastructure

For the management of infrastructure over long periods of time and for the assessment of infrastructure that is already deteriorating, structural engineers need to be able to predict the likely future deterioration of their structures. Conventionally they rely on mathematical models to do this. This paper deals with the development of such models for the marine corrosion of steel. They are based on corrosion science fundamentals and microbiological aspects. They invoke probabilistic notions to account for the inevitable uncertainties attached to enumerating the environment and other conditions and in prediction. Observations of long-term data, interpreting it and then invoking intuition and scientific principles are all involved in developing models consistent with expectations and with boundary conditions. They are then calibrated to data from the literature and using new data specifically collected as required. The approach taken herein for model development focuses on interpreting data in terms of an ‘a priori’ model in the sense of Bayes inferential theory, rather than using data alone to dictate a ‘model’, such as typically resulting from correlation studies. Some results obtained to-date illustrate that this approach can be successful in providing new interpretations and better understanding of empirical ‘field’ observations. A brief outline is given also of future research directions.