Computing Systems in Engineering最新文献

A cell center, upwind biased, spatially third-order and temporally second-order finite volume procedure for solving the Maxwell equations in time domain has been ported to a Intel Touchstone Delta and a Paragon multicomputer. Using a one-dimensional domain partition scheme, the computer code has attained a data processing rate of 7.56 Gigaops on the 512 nodes of the Delta and 3.56 Gigaops on the 179 nodes of the Paragon XP/S systems. The high parallel efficiency of the present computer program however, is sustainable only up to a limited size of addressable memory (nodes × 48 × 96). The scalable performance range of the present code is significantly extended by operating on the Paragon system at the present time.

Research efforts aimed at optimizing the computational efficiency of the p-method are described. These include (i) a novel quadrature scheme for hierarchical shell elements, (ii) a family of assumed strain hierarchical shell elements, (iii) selective polynomial order escalation for assumed strain elements, and (iv) accelerated multigrid-like solution procedures. Numerical experiments indicate that with these enhancements it is possible to speed up the overall computational time of p-method for analysis of shells by a factor greater than five for relatively small problems (less than 10 000 dofs), while computational savings for larger problems are even more significant. It has been found that the performance of the enhanced variant of the p-method for shells is comparable to that of the h-method for low accuracy requirements, and better if higher accuracies are desired.

A computational scheme is introduced for the integration of rate-dependent multiple-slip crystal plasticity constitutive relations. Fundamental issues of accuracy, stability, and stiffness that are intrinsically related to the evolution of microstructural failure modes in metallic crystals are addressed. An adaptive finite-element methodology is introduced to classify these characteristics. A nonlinear initial value system is derived to update the plastic deformation-rate tensor. An explicit method is used in non-stiff domains, where accuracy is required. If a time-step reduction is due to stability, a harbinger of numerical stiffness, the algorithm is automatically switched to an A-stable method. A stiffness ratio is defined to measure the eigenvalue dispersion of the system. The adaptability of the proposed algorithm for the solution of a class of inelastic constitutive relations is illustrated by investigating the influence of high angle grain boundary orientations on failure in face-centered cubic (f.c.c.) bicrystals. The effects of grain boundary misorientation, dislocation densities, strain hardening, and geometrical softening on failure evolution are investigated. This study underscores the importance of understanding the origin of numerical instabilities, such that these instabilities are not mistaken for inherent material instabilities.

This paper describes the development of an expert system prototype for forecasting contract cost and duration of housing modernization works in the U.K. There are two main phases in the system's operation: the first for generating a plan for a given project, and the second for simulating construction progress for cost-time forecasting. A probabilistic approach was chosen using the Monte-Carlo method for establishing productivity rate. If and when interference or delay occurs during the simulation, the decision is made with the help of integrated in situ knowledge. Visualization status provides a tool for intelligent simulation of the construction phase of the work. An object-oriented programming approach for carrying out these aims is described. The system and its knowledge base structure are presented. Finally, conclusions are drawn.

A new mapping algorithm is presented for domain decomposition for the purpose of allowing researchers to conduct finite element analysis on massively parallel computers. Over the last few years, massively parallel MIMD machines such as the Intel Touchstone Delta and recently the Intel Touchstone Paragon have become increasingly popular for speeding up finite element computations. Most of these applications use domain decomposition as a first step towards conquering the problem. Many different algorithms have been developed by researchers to achieve an effective domain decomposition. Some of these methods use connectivity information only, some use coordinate information only, while others use both of them together. Some algorithms are based on assigning weights to nodes using a particular strategy while others are recursive in nature. As will be discussed in this paper, the logic employed in various algorithms works perfectly well for certain meshes to be decomposed, in certain numbers of subdomains; while it gives far from perfect results for other meshes or for same meshes to be decomposed in a different number of subdomains. The logic used in the proposed algorithm has been developed in a creative way such that it is closer to a human's natural thinking when making decisions. Fairly large meshes can be decomposed in a matter of seconds on a Sun Sparc station by the proposed algorithm. Its execution time remains almost the same for any number of subdomains.

Problem decomposition strategies for load balancing parallel computations on adaptive hp finite element discretizations are discussed in this work. The special difficulties that arise in partitioning these discretizations are highlighted. Three classes of algorithms—mesh traversal based on orderings, interface based decompositions and recursive bisection of orderings are discussed. A new ordering scheme for efficient recursive bisection of orderings is introduced. Details of the algorithms and examples along with discussions of their merits and demerits are presented. Recursive bisection on the new ordering introduced here outperforms several known algorithms on test cases.

We consider the problem of modeling very complex physical systems by a network of collaborating PDE solvers. Various aspects of this problem are examined from the points of view of real applications, modern computer science technology, and their impact on numerical methods. The related methodologies include network of collaborating software modules, object-oriented programming and domain decomposition. We present an approach which combines independent PDE solvers for simple domains which collaborate using interface relaxation to solve complex problems. The mathematical properties and application examples are discussed. A software system RELAX is described which is implemented as a platform to test various relaxers and to solve complex problems using this approach. Both theory and practice show that this is a promising approach for efficiently solving complicated problems on modern computer environments.

The Design Sensitivity Analysis and Optimization (DSO) tool, developed initially for sizing design application, has been extended to support shape design applications of structural components. The new capabilities including shape design parameterization, error analysis and mesh adaptation, design velocity field computation, shape design sensitivity analysis, and interactive design steps, are discussed. These capabilities are integrated on the top of the DSO framework that includes databases, user interface, foundation class and remote module. The DSO allows the design engineer to easily create geometric, design, and analysis models; define performance measures; perform design sensitivity analysis (DSA); and carry out a four-step interactive design process that includes visual display of design sensitivity, what-if study, trade-off analysis, and interactive design optimization. Additionally, a 3-D tracked vehicle clevis is presented in this paper to demonstrate the new capabilities.

In this paper an automated approach is used to carry out sensitivity analysis and to obtain optimum shapes for plates and shells in which the natural frequencies are maximized. The free vibration analysis is carried out with the nine-noded, degenerated, Huang-Hinton shell element implemented and tested in Part I of this paper. Design variables that specify either the shape or thickness distribution of the structures are considered. Special attention is focused on the sensitivity calculations and problems connected with their accuracy and performance are highlighted when the semi-analytical and finite difference methods are used. Advantages and disadvantages of each method are discussed. The optimal solution is found by the use of a structural optimization algorithm which integrates the finite element module (Part I), sensitivity analysis and a mathematical programming method: sequential quadratic programming (SQP). Optimal forms are then obtained for a set of benchmark examples using the two sensitivity analysis techniques and their results are compared. The results obtained for optimum solutions in the present paper justify the usage of the semi-analytical method for sensitivities calculations for structural shape optimization purposes.