Owing to the failure of the finite element analyst to employ a properly refined computational model, the accuracy of preliminary finite element computations is often low. Thus, it is useful to introduce a computational procedure for improving the results obtained from a preliminary finite element solution. Such a procedure is presented in this paper. In this procedure the solution error components are decomposed into two parts. One of the error components is assumed to have a long period variation. The other component is assumed to have a short period variation. Multiplicative and additive correction procedures are introduced to iteratively eliminate the two error components. The multiplicative and additive correction procedures are implemented using multi‐level solution techniques.
|頁（從 - 到）||27-41|
|期刊||International Journal for Numerical Methods in Engineering|
|出版狀態||Published - 1989 一月|
All Science Journal Classification (ASJC) codes
- Numerical Analysis
- Applied Mathematics