Search results

This paper presents the development of a parallel finite element algorithm for a MIMD parallel computer. The elements are distributed, onto the processors, in such a way that neighbouring elements are placed onto neighbouring processors. This guarantees a good load‐balancing even in physically non‐linear computations. The distribution of the columns of the global system of equations is done by an election algorithm. For solving the global system of equations we use a parallel preconditioned conjugate gradient solver. Tests were done with an elastoplastic material model for proving the efficiency of assembling and solving the system of equations.

The boundary element method (BEM) and the finite element method (FEM) may be computationally expensive if complex problems are to be solved; thus there is the need of implementing them on fast computer architectures, especially parallel computers. Because these methods are complementary to each other, the coupling of FEM and BEM is widely used. In this paper, the coupling of displacement‐based FEM and collocation BEM and its implementation on a distributed memory system (Parsytec MultiCluster2) is described. The parallelization is performed by data partitioning which leads to a very high efficiency. As model problems, we assume linear elasticity for the boundary element method and elastoplasticity for the finite element method. The efficiency of our implementation is shown by various test examples. By numerical examples we show that a multiplicative Schwarz method for coupling BEM with FEM is very well suited for parallel implementation.

The finite volume method for radiative heat transfer calculations has been parallelized using two strategies, the angular domain decomposition and the spatial domain decomposition. In the first case each processor performs the calculations for the whole domain and for a subset of control angles, while in the second case each processor deals with all the control angles but only treats a spatial subdomain. The method is applied to three‐dimensional rectangular enclosures containing a grey emitting‐absorbing medium. The results obtained show that the number of iterations required to achieve convergence is independent of the number of processors in the angular decomposition strategy, but increases with the number of processors in the domain decomposition method. As a consequence, higher parallel efficiencies are obtained in the first case. The influence of the angular discretization, grid size and absorption coefficient of the medium on the parallel performance is also investigated.

Describes a European project to develop vision‐based, closed‐loop robot control technology to facilitate assembly tasks in the automotive industry. Looks particularly at the problem of placing a sunroof on a car body accurately.

Information Processing Model of Knowledge Systems Two Massachusetts Institute of Technology (MIT) scientists, Kaan Egilmez and Steven H. Kim, writing in Applied Artificial Intelligence, Vol. 5 No. 3, 1991, pp. 353–92, believe that to date, knowledge‐based systems (KBSs) have been developed on an ad‐hoc basis, without the benefit of quantitative tools for analysis and synthesis. They claim that to approach the field of knowledge engineering systematically, we must develop performance measures to characterize the effectiveness and efficiency of knowledge systems, and thereby serve to highlight directions for improvement. They say that, when KBSs are applied to control tasks, real‐time performance becomes critical and in consequence the internal processes of knowledge‐based control systems (KBCS) must be optimized to fulfil real‐time requirements. In their contribution they derive relationships between information attributes and system performance.