When I do a Frequency Domain Decomposition (FDD) I always get mode shapes with significant imaginary components.
Let's say I see a strong modal response. Still have strong complex nature.
When I do a simple PCA of the dataset I can "see" the same shape in the dominant mode, but obviously no imaginary component.
So I interpret the complex values from the FDD as basically being a better fit to the data, with a phase component being permitted.
- Is that interpretation accurate?
- Is there a good way to go from the FDD shape to a non-complex shape? Taking the real doesn't work, nor the magnitude (all positive)?
Should I just be using the FDD to identify frequencies of interest, bandpass filter before PCA?