Published online by Cambridge University Press: 28 February 2011
The prediction of the long term corrosion of metal containers for nuclear waste under geological disposal conditions requires the extrapolation of corrosion data over several hundred years. There is a general trend to use either simulation or ‘worst case’ experiments as a means of materials selection, but neither really addresses the problem of long term extrapolation. It is proposed that such an extrapolation can only be done convincingly if it is based on a sound and generally accepted mechanistic understanding of the processes involved. If such knowledge does not exist the first step must be to acquire it through experimental mechanistic studies. Subsequently such knowledge should be formulated into mathematical models, which can be used to make long term predictions, and which can be validated by comparison with short term experimental results. The application of this combined mathematical modelling/experimental approach is illustrated for three corrosion processes which may affect carbon steel containers, namely general corrosion, localised attack and stress corrosion cracking.