Principal component analysis on 3D scanned compressor blades for probabilistic CFD simulation
Alexander Lange, Matthias Voigt, Konrad Vogeler, Erik Johann
abstract:
The Principal Component Analysis (PCA) is a widely applied method for describing geometric variability of turbo-engine airfoils, where a set of correlated measured coordinates is converted into a set of mode shapes and uncorrelated modal amplitudes. The presentation discusses the required steps to apply this method to 3D scanned compressor blades. Transformation laws are derived for translating the unstructured scan into the analysis mesh for performing the PCA and also for translating geometric mode shapes from the analysis mesh to the CFD mesh. Several sampling technologies are discussed concerning their applicability for correctly modelling the statistics of the modal amplitudes matrix. The mentioned points were integrated into a process chain that is applied to a high-pressure compressor stage (HPC) of a jet-engine represented by a stator-rotor-stator configuration. Within a Monte-Carlo simulation (MCS) 500 probabilistic realizations of the compressor blade were created and the impact of the manufacturing variability on the performance of the HPC stage is investigated. The post-processing of the probabilistic simulation reveals the sensitivity of the HPC stage performance on the individual mode shapes and derived geometric features.
reference:
Alexander Lange, Matthias Voigt, Konrad Vogeler, Erik Johann
"Principal component analysis on 3D scanned compressor blades for probabilistic CFD simulation"
14th AIAA Non-Deterministic Approaches Conference, Honolulu, Hawaii, 2012.