To control the numerical quality of non-linear simulations allows to automate the choice of discretisation. Within the framework of time dependent problems, this quality depends on the error made in the calculation of the transformation history. Generally, before obtaining a calculation of a satisfactory quality, it is necessary to carry out an evaluation on the numerical approximations to use by exploiting all the history of approximate solution defaults. We propose a strategy which naturally includes this expert stage. The desired solution is built iteratively, bettering all the history at each iteration thanks to the LATIN method. A global error estimator is provided by the dissipation error. To pilote the adaptive computation, we propose a partition of this estimator in indicator associated to each kind of error contribution.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £75 +P&P)