ABSTRACT:
This paper describes an unstructured macro-cell (UMC) algorithm
for solving partial differential equations, and its application
to computational fluid dynamics. A macro-cell is an intermediate-sized
domain, consisting of roughly 10 or more grid points in each spatial
direction in a structured array. The governing equations are solved
on this structured grid within a macro-cell. A complete computational
domain is composed of perhaps 10s to 1000s of macro-cells which are
arranged in an unstructured manner. Local grid refinement is
achieved by splitting a macro-cell, recursively as needed, in a
manner similar to that of adaptive mesh refinement (AMR) schemes.
The UMC algorithm has features of structured and unstructured
grid methods, and it is intended to retain the benefits and avoid
the drawbacks of purely structured and unstructured methods.
Because the equations are solved on a structured array of
grid points, high-accuracy finite-volume and implicit methods
can be used. The use of unstructured macro-cells enables the
use of several geometry- and solution-adaptive strategies.
Thus, a UMC method offers many significant benefits for CFD.
The principal difficulty which must be overcome is the need
for methods of achieving high accuracy at macro-cell boundaries. This
paper addresses the general concepts of the macro-cell solver
with emphasis on boundary issues. Companion papers by
Treidler et al. (1999) and Reisenthel and Childs (1999) describe
aspects of the computational efficiency and the solution-adaption
algorithm, respectively.
ABSTRACT:
Three local spectral estimation schemes were compared for
their ability to accurately detect the presence of excess
energy in high frequency modes lying outside the accuracy
bandwidth of a given discretization scheme: the windowed
Fast Fourier Transform (FFT), the Discrete Wavelet
Transform (DWT), and the Maximum Entropy Method
(MEM). Systematic deterministic and statistical tests were
carried out, indicating the MEM to be the most accurate.
An MEM-based on-the-fly grid refinement / derefinement
scheme was developed. Its capabilities are demonstrated in
terms of local grid refinement, local grid coarsening, and
dynamic adaptation.
ABSTRACT: Preliminary results are presented from application
of implicit integration schemes to high-accuracy CFD methods.
High-accuracy refers to spatial discretization methods which
are optimized for spectral bandwidth, rather than order of accuracy
based on a Taylor series expansion. The results show that reductions
of more than 90% in computational effort, as measured by CPU time,
can be achieved. The reduction is relative to second-order methods
for obtaining steady solutions to the same level of accuracy. In
addition, a variety of time integration methods have been evaluated
for their use in time-accurate and steady-state simulations with the
high-accuracy central difference schemes.
CFD-Based Multidisciplinary Design Optimization:
ABSTRACT: The objective of this study was to demonstrate the
feasibility of integrating modern CFD aerodynamic prediction
technology in ASTROS. A detailed analysis is presented of how
to make use of both steady and unsteady CFD-generated air loads
in the design disciplines of ASTROS. In the first part of this work,
two Aerodynamic Influence Coefficient (AIC) alternatives were
evaluated. The first one is based on the Direct Iterative Surface
Curvature (DISC) method, which relates a local perturbation in
surface curvature to a corresponding pressure perturbation. The
second alternative was the use of CFD automatic code differentiation
to compute sensitivities to geometric design variables. In the
second part of this work, a demonstration of a Computational Fluid
Interface was carried out by coupling ASTROS with the transonic small
disturbance computational aeroelasticity program CAPTSD. This
demonstration involved the redesign of a composite wing under
transonic load conditions, subject to static aeroelastic
constraints. The results demonstrate that it is possible to
successfully couple the design disciplines of ASTROS with Computational
Fluid Dynamics codes, and point to possible improvements of the trim
module for successful integration with CFD.