PSI - Issue 57

Laurent Dastugue et al. / Procedia Structural Integrity 57 (2024) 355–364 Michael Klein et. al./ Structural Integrity Procedia 00 (2019) 000 – 000

361

7

space is required. This also saves manual care of the files and possibly copying to the right place. If necessary, all stress results can be output on file optionally.

Fig. 10. “On -the- fly” means no stress data on files.

To illustrate the required storage space an example with small model size shows the need for the data efficiency to save the storage of stress data. The small example (Fig. 11) has only 1573 nodes for which 1573 stress tensors must be saved in a file on disk. If we assume 10,000 timesteps of a measurement from a torture track, 403 MB are necessary to save the stress data in a binary HDF (hierarchical data format) file. But realistic models are larger by a factor of 1000 and the number of time steps by a factor of 10. If these factors are considered, the stress tensors would require 4.03 TB. The required write and read times are not acceptable for industrial tasks. The required disk space exceeds the typical capacity, which makes it impossible to use fatigue analysis for such models.

Fig. 11. Small example structure.

By using “On -the- fly” calculation approach the stress tensors are n ot written to disk. The file storage issue is completely resolved. Outcome of the data efficiency is that former impossible model sizes are now possible. 4. Available methods in integrated fatigue life analysis The fatigue life analysis integrated in PERMAS allows analyzes for high cycle fatigue to be carried out. The load data for such a calculation are nodal point stresses from a finite element (FE) analysis. In principle, damage values at FE nodes are calculated from a stress-time profile in a fatigue life analysis. The stress data are node stresses or stresses by superconvergent patch recovery (SPR) procedure. Optionally, the associated stress gradients can be used for the calculation of support numbers for both types of stress.