Keynote Lecture: Rob Melchers

The quality of life-cycle assessments and civil and structural engineering reliability analyses depends much on the quality of the models on which they are based. In general, models for loading systems and for properties of materials are well-established and have well-defined probability distributions. Models for structural or material deterioration, however, are not in the same class. In this paper some examples are given where the deterioration models proposed were obtained from assumed behaviours without critical or close examination of the underlying physico-chemical fundamentals and requirements and the available, often only statistical, and perhaps uncertain data. It is emphasized that there must be good understanding of the underlying principles to obtain models that are not entirely empirical and that have some degree of scientific validity. In addition, experimental laboratory data must be used with utmost caution as it seldom represents long-term behaviour and data that can sometimes be obtained from field experiments. Some examples of these aspects are given.

Further, it is demonstrated by example that careful use of statistical data and its analysis can add value to the modelling effort. The paper describes modelling of corrosion processes applicable to steel, aluminium and copper-nickels, the interpretation of uncertain (statistical) field data and the way that data and understanding can sometimes give rise to completely new interpretations of (part) of the underlying physico-chemical processes, or allow differentiation between scientific observations and practical applications. The latter is illustrated for stainless steels and for cast iron pipes buried in soils.