Related Posts

Ansys R2

Linear Dynamics Just Got a Whole Lot Faster

Ball Grid Array (BGA)

Do you routinely tackle mechanical simulations in the field of linear dynamics? Are you unhappy with your simulation times? If you answered yes to both of these questions then keep on reading. ANSYS Mechanical 17.0  has many exciting new features — far too many to cover here. This post focuses on only one of these new features: the improved performance for linear dynamics simulations in Distributed ANSYS.

The performance improvements mentioned here were not achieved overnight, or even a single release. Rather they were a culmination of a series of projects involving several years of effort and spanning several releases. In ANSYS 16.0, support was added in Distributed ANSYS for the mode superposition method with harmonic/transient analyses. In  ANSYS 17.0, spectrum analyses — including response spectrum analyses (SPRS and MPRS) and random vibration analyses (PSD) — were added to the list of supported analysis types in Distributed ANSYS. The past few releases have also seen the introduction of six eigensolvers, which all run fully parallel, in Distributed ANSYS. These include the popular Block Lanczos and Subspace algorithms, along with QRDAMP, UNSYM, DAMP, and PCG Lanczos eigensolvers.

But of course we cannot discuss performance improvements without the obligatory chart. So here it is.

This chart reveals the substantial speedups obtainable now that the entire workflow for linear dynamics simulations is parallelized in Mechanical. The solver rating (i.e., number of simulations per day) is plotted as a function of the core count. For this chart, a simulation of an engine comprising of 3.7 million degrees of freedom is solved using several releases. To obtain the dynamics of the engine, a modal analysis requesting 250 modes is solved followed by a harmonic analysis using the mode superposition method. In past releases, the mode superposition calculations that occur after the modal analysis involved virtually no parallelism. That meant the bulk of the harmonic analysis time was spent number crunching on a single CPU core. Extremely inefficient!

In ANSYS 16.0, these calculations were parallelized resulting in an 11x speedup over previous versions. In ANSYS 17.0, a new parallel version of the Block Lanczos eigensolver is introduced that significantly speeds up the calculation of the natural frequencies during the modal analysis. This further extends the gains such that version ANSYS 17.0 is an amazing 26 times faster than ANSYS 14.5!

Learn More

Learn more about ANSYS Mechanical.