If you have any feedback on how we can make our new website better please do contact us. We would like to hear from you.
Fast solution of combined partial differential equations (PDEs, Navier-Stokes, Maxwell, Fourier, etc.) is one of the fundamental tools for modern and future engineering and scientific applications. Fast simulations are used for design validation and verification. If the simulation code can run in real time (the simulation have the same response time of the physical phenomena), the model can be used to improve the closed loop controller and increase the reliability of the system using built-in failure detection algorithms.

The state of the art of the PDE simulations is particularly disappointing one. Most of the actual numerical methods comes from the single-cpu era. Staring from the sixties, computer architecture have tried to reduce the simulation time using parallel vector operations (SIMD, single instruction, multiple data). A notable examples were the CDC an CRAY computers. Present GPUs uses the very same architectures. Modern workstations uses parallel, independent processors (MIMD, multiple instruction, multiple data) sharing a common memory pool. Present time supercomputers are combinations of SIMD and MIMD computing elements, shared memory pool (DRAM) interconnected using fast (1-100 Gbit/s) network.
Despite these improvements, PDE simulations are painful slow respect the real world physical phenomena.  

Model based real time control with failure detection

FRINGE: an outer edge; margin; periphery; something regarded as peripheral, marginal, secondary, or extreme in relation to something else. In a world where uncertainty, doubt and indetermination are the dominant factors, unconventional solutions are the way: the unthinkable becomes possible, the unfair becomes standard and the impossible, inevitable.
  Site Map