Validity and Quality of Deterioration Models for Structural Reliability Assessment

There are increasing pressures to extend the service lives of existing infrastructure and to provide optimal management of their maintenance. Structural reliability theory now provides a very comprehensive approach to assessing risks for complex infrastructure systems. It requires probabilistic models for the predic- tion of the long-term changes to structural capacity and resistance as may result, for example, from changes in the management or operation of the asset or from material deterioration in harsh operational environments. While empirical mod- els may be sufficient in some cases, it is preferable for the models to be based on fundamental science and to deal with remaining uncertainty in a probabilistic man- ner. Herein models for the prediction of the corrosion of structural steel in marine environments are considered. The issues involved are reviewed and some of the pitfalls in model development are considered. These include (i) models based on inadequate underlying theories, (ii) use of inhomogeneous populations, (iii) use of inappropriate influencing factors, (iv) use of laboratory test results rather than field results, (v) insufficient attention to limitations and conditions, (vi) insufficient understanding of the changes in conditions with time, and (vii) invoking model simplifications too early. Drawing on material that has already been published in the literature, examples are given of each to show how these matters may provide misleading information to potential end-users.