Prediction of helicopter stability in trimmed flight is often based on Floquet theory. It involves a non-linear trim analysis for the control inputs and corresponding periodic responses, and then a linearized stability analysis for the Floquet transition matrix (FTM), and eigenvalues and eigenvectors of this matrix. The shooting method with damped Newton iteration is used for the trim analysis and generates the FTM as a byproduct. The QR method is used for the eigenanalysis. The Floquet analysis comprises the shooting and QR methods. A rotor with Q identical and equally spaced blades has Q planes of symmetry. The fast-Floquet analysis exploits this symmetry, and thereby provides nearly a Q-fold reduction in run time and frequency indeterminacy of the Floquet analysis. Still, the run time for the fast-Floquet analysis on serial computers becomes prohibitive for large models (order or number of states > 100); in fact, it grows between quadratically and cubically with the order, and the bulk of it is for the trim analysis. Accordingly, a parallel fast-Floquet analysis is developed for both types of mainstream parallel computing hardware: distributed computing systems of networked workstations and massively parallel computers; algorithmically, both types belong to the MIMD (Multiple-Instruction, Multiple-Data) architecture. Large models with hundreds of states are treated, and a comprehensive database on parallel performance and computational reliability is generated. Computational reliability is quantified by parameters such as the eigenvalue condition number. Similarly, parallel performance is measured by parameters such as speedup and efficiency, which collectively provide a measure of how fast a job can be completed with a set of processors and how well the processors are utilized. Compared to the serial fast-Floquet analysis, the parallel analysis reduces the run time dramatically and provides a practical means of controlling the growth of run time with the order by a judicious combination of speedup and efficiency.
Fast-floquet analysis of helicopter trim and stability with distributed and massively parallel computing
Floquetsche Schnellanalyse der Hubschraubertrimmung und -stabilität mit dezentralen und massiv parallelen Rechnern
1997
17 Seiten, 10 Bilder, 1 Tabelle, 17 Quellen
Conference paper
English
British Library Conference Proceedings | 1997
|Parallel Fast-Floquet Analysis of Trim and Stability for Large Helicopter Models
British Library Conference Proceedings | 1996
|Parallel Computing Concepts and Methods for Floquet Analysis of Helicopter Trim and Stability
Online Contents | 1996
|British Library Conference Proceedings | 1993
|