Abstract

This effort investigates the ability to efficiently emulate the modal characteristics of a transonic compressor rotor as a function of measured geometric variations. The transonic rotor is measured with a structured blue light scanner that provides a dense geometric representation of the as-manufactured part. Principal component analysis (PCA) is applied to create a reduced order, orthogonal basis of parameters used as independent emulator parameters. Ordinary least squares (OLS) and Gaussian stochastic process (GSP) regression models are employed as emulators for airfoil frequency and mode shape across the first twenty resonant modes. While many modes are emulated with high accuracy, some challenge conventional emulation. It is shown that geometric variations cause significant variation in modal behavior for closely spaced frequencies. This work identifies that as-manufactured geometry deviations can lead to frequency veering and associated chaotic modal behavior. Effective emulation of frequncy and mode shape is demonstrated, but the physics-based understanding of as-manufactured geometry frequeny veering defines the way for future improvements in emulation accuracy and computational requirements.

This content is only available via PDF.
You do not currently have access to this content.