PSI - Issue 75

Laurent Dastugue et al. / Procedia Structural Integrity 75 (2025) 334–343 Laurent Dastugue et al. / Structural Integrity Procedia (2025)

337

4

through better integration and workflow optimization, ultimately enabling fatigue analysis to become a native part of the simulation pipeline. 3. New Approach – Full Integration Historically, fatigue analysis has been performed using a workflow as it described in chapter 2. In this, the classical setup (see left image of fig. 4), the finite element analysis (FEA) solver computes stress results, which are then exported — typically via large, intermediary result files — for use by an external fatigue solver (post-processor). This approach introduces several inefficiencies: significant I/O overhead, disk space consumption, and the potential for manual mistakes in the workflow. To address these limitations, various semi-integrated solutions (middle image of fig. 4) have been developed. These involve tighter coupling between structural and fatigue analysis software, reducing the need for user-driven file management and simplifying the overall workflow. However, such solutions often rely on internal APIs or intermediate data exchange layers, and therefore maintain a fundamentally modular architecture. While usability is improved, core issues — such as data duplication, limited access to solver internals, and suboptimal performance — remain unresolved.

Fig. 4. Integration levels: classic, semi-integration and full integration.

The proposed fully integrated fatigue analysis (see right image of fig. 4) approach overcomes these constraints by embedding fatigue computation natively within the FEA solver. This architecture eliminates the need for external interfaces or data export. All fatigue-relevant computations are performed within the same process, using a shared, internal data model. As a result, the workflow becomes significantly more efficient: disk I/O is minimized, the need for redundant storage is eliminated, and the handling of simulation data is greatly simplified. Crucially, the integrated implementation allows the fatigue module to access the complete set of internal solver data structures. Since the solver is inherently designed for high-performance computing (HPC) environments, it already manages data in a highly parallelized and memory-optimized manner. Fatigue evaluation can directly utilize existing data such as: • Element types and connectivity • Stress tensors at Gauss points or other integration locations • Stress gradients Within this framework, fatigue damage becomes a secondary result type, analogous to stress or strain. The additional computational cost is minimal due to the reuse of already available data and infrastructure. However, achieving this level of integration necessitates a complete re-implementation of the fatigue algorithm within the FEA solver environment. Existing fatigue solvers, originally designed as stand-alone tools, cannot simply