Carassale, L. & Vernazzani, A. (2009). 5th European and African Conference on Wind Engineering - 5EACWE, Florence.
Abstract
A common activity in
experimental bluff body aerodynamics is the measurement of pressure fields
acting on models by multi-channel scanners. These measurements are often
analyzed by multi-variate statistical techniques such as Principal Component
Analysis (PCA), more commonly referred as Proper Orthogonal Decomposition
(POD), which represents the measurements as a linear combination of
deterministic vectors, the PCA modes, modulated by uncorrelated amplitudes. The
PCA modes have often been interpreted as elementary pressure patterns, whose
characteristics reflect the aerodynamic behavior of the model, and used for
interpretation purpose. However, a strong limitation involved in the use of PCA as a pattern
identification tool is related to the orthogonality of its modes that, from a
physical point of view, is not justifiable. With the purpose of obtaining a
representation formula analogous to PCA, but based on a non-orthogonal set of
modes, the concept of Independent Component Analysis (ICA) is briefly described
and applied to the analysis of pressure measurements carried out of a high-rise
building model.
|
