A new simpler solution procedure is presented for the finite element analysis of creep problems. The creep strains are eliminated as computation variables. At each time step, a system is solved for the stresses and velocities alone.
Improved sequentially linear solution procedure
✍ Scribed by Jan Eliáš; Petr Frantík; Miroslav Vořechovský
- Publisher
- Elsevier Science
- Year
- 2010
- Tongue
- English
- Weight
- 936 KB
- Volume
- 77
- Category
- Article
- ISSN
- 0013-7944
No coin nor oath required. For personal study only.
✦ Synopsis
The article proposes an improvement over the widely used sequentially linear solution procedure often utilized for fracture simulations. In the classical secant version of this method, a partial solution of a step is scaled to reach a stress limit in exactly one element and the mechanical properties of the critical element are reduced. Non-proportional loading is generally unfeasible due to avalanches of ruptures caused by stress redistribution. Because only one loading vector can be scaled at a time, all others have to remain constant during the step. However, the constant load vectors do not allow proper determination of the critical element. A modified procedure based on redistribution of released stresses is developed here. It preserves the linearity of each step. After rupture of the critical element, a sequentially linear redistribution process of stress release takes place until a static equilibrium state is reached. During the redistribution, other elements may break.
The proposed enhanced sequential procedure is also compared with another recently published ''event-by-event" linear method for non-proportional loading. It is shown here, with the help of simple examples, that the proposed redistribution method yields correct results for non-proportional loading, unlike the other methods under comparison.
📜 SIMILAR VOLUMES
When a physical system, or its topological model, is torn apart into n independent parts, a factorized form of the true inverse matrix Z may be established by finding the inverse of n + 1 much smaller matrices. No additional matrix multiplications are needed. The first n matrices Zr represent the n