ABSTRACT
The development of parallel computational methods, including parallel numerical algorithms and parallel processing techniques, has become a necessity and has generated new challenges in structural mechanics. Classic structural formulating methods and numerical methods are generally unable to exploit multiprocessors and powerful hardware. This chapter extends from developing parallel processing techniques for the finite element method to discovering inherent parallel numerical algorithms that make them attractive bases for parallel computation. Element-by-element solution strategy also shows potential application on parallel computing systems. Much of the computational effort of the finite element method in a large structure involves the solution of a system of linear algebraic equations. The global stiffness matrix generated from this system of linear algebraic equations is symmetric, positive, definite, and banded. Parallel backward substitution starts subsequently in a similar logical process but in the reverse row order.
