Next: Polymeric Fluids
Polymeric Fluid Dynamics
on Parallel Supercomputers
David S. Malkus and
Constitutive properties of
Polymeric (non-Newtonian) Fluids
has been a topic for many researchers in the last several decades.
Understanding polymer fluid dynamics is important for the industry in connection with
plastics manufacture, performance of lubricants, application of paints,
processing of foodstuffs, and movement of biological fluids.
Numerical schemes have been developed to match the experiments both
qualitatively and quantitatively. One of the most common examples to test
the numerical schemes is the contraction flow problem. Although the
cause of numerical breakdowns remain a controversy, one approach
is to include the spurt phenomena in the model, which requires that
fully dynamic equations be considered. Solution of a non-Newtonian flow
is computationally intensive because of the nonlinearity both in
the equation of motion and in the constitutive equations. Also, solving
fully dynamic equations is time-consuming due to the small time steps.
We describe a parallel finite element algorithm and its implementation on the
Intel architectures to tackle this problem.
Parallel versions (on 16 nodes) run almost 10 times faster than the
serial version on Sun SPARCstations. The parallel efficiency has been
measured to be more than 70 % and the performance on 16 nodes is 2
times better than the CRAY Y-MP version.
Tue Jan 21 16:43:41 EST 1997