you are here:   / News & Insights / Engineering Advantage Blog / Capturing Time History Results from Large Scale Finite Element Analyses for Design

Engineering Advantage

Capturing Time History Results from Large Scale Finite Element Analyses for Design

November 14, 2013 By: Peter Barrett

Stress analysis engineers can now model complex nonlinear dynamic finite element analyses of very complex structures using distributed processing on a desktop computer.  However, it is very difficult to isolate structural design data from the terabytes of analysis results produced in a single analysis run.

Example analyses include: the seismic analysis of a nuclear containment building under uplift loading, the complex range of motion of a hip or knee prosthesis, and the fluid-structure interaction loads on the walls of a nuclear spent fuel pool.  The design team wants peak stresses for design of critical sections, while the analysis team barely has enough disk storage to capture just the displacement time histories. Storing all stresses and strains, for example, could require multiple terabytes of disk space for each analysis run.

A common approach to isolate data for design is to utilize submodels.  In typical submodeling, a finer mesh is used to create an independent finite element model where the boundary displacements are mapped from the global analysis.  The submodeling approach proposed herein reuses either the full global model or a subset of the global finite element model in the areas of interest and only stores peak design data.   

The local sector is isolated and each time point is solved in a series of independent static analyses where displacements are prescribed for every degree of freedom from the full dynamic global analysis.  Since the identical finite element model is used, and all degrees of freedom are prescribed, the modeling and analysis time is negligible.  Results data is limited to a single set of results.  The first set of results are solved for and critical design stresses and strains are copied to an external array.  After each subsequent solution, standard postprocessing routines again extract the peak values and compare with the values stored in the external array.  Existing results are replaced with the new peak values, as required.  The process is repeated for all the global model time steps.  Upon completion of the sequence, a single array of results provides the envelope of peak values required  for design.  The amount of data storage required in the submodel never exceeds a single load step set of results and thus can easily be transferred to the design engineer for detailed local design.