In every industry, products are becoming more complex, and thus simulation must become more detailed and comprehensive in order to predict their behavior. Whether that product is a car or a blender, the key to understanding its performance – before a physical model is fabricated – is creating digital continuity, in a virtual environment, with the actual behavior that is being predicted.
This means that a quality simulation product is key. The closer the simulation can come to the real world behavior of the product, the better chance the product has of performing in accordance with expectations in a real world environment. In order to better predict real behavior through simulations, engineers require detailed discretized models to accurately capture product behavior, as well as physics solvers that can accurately capture material behavior.
These requirements have led to some great enhancements in simulation technology. Rather than detailed models of individual components, simulation software is presenting models of full scale systems and subsystems. These large-scale simulation models are the result of better solution fidelity and fewer modeling abstractions. In the last decade and a half or so, model sizes have shifted from about 100,000 degrees of freedom (DoFs) to hundreds of million DoFs.
As model sizes get bigger, it becomes more expensive to utilize a traditional direct solver, so iterative solvers are moving in to complement them. Iterative solvers are generally faster, require less memory, offer scalable performance and are relatively robust for well-conditioned problems. In addition to DoFs, iterative solvers can also handle more Floating Point Operations, or FLOPs.
Abaqus now offers a new iterative solver, also available on the 3DEXPERIENCE platform. It is AMG-based, which stands for algebraic multi-grid. A multi-scale resolution approach results in fast performance. Between different scale resolutions, numerical methods – restriction, prolongation, and smoothing – are being used. New aggregation and smooth algorithms were developed in order to achieve a faster, more robust performance, resulting in an iterative solver that is suitable for large problems.
The new AMG-based iterative solver was developed with FE applications in mind. It is scalable, robust and can support a large number of simulation models, including contact, gaskets, pre-loadings, distributive and kinematic couplings, periodicity boundary conditions, significantly varying material properties, and poorly shaped elements.
The iterative solver addresses customer demand for high-fidelity and high-accuracy virtual models. It enables the solving of extremely large simulation problems – up to approximately 200 million DoFs – with excellent performance and scalability and relatively low costs.
Repost from Simulia’s blog.