Spectral Analysis

Introduction

A physical process can be described in the time domain (amplitude h(t) as a function of time) or the frequency domain (amplitude H(f) as a function of frequency). The time-domain view gives sample amplitudes at each sampled instant; often, knowing a signal's frequency content is more useful than knowing individual sample amplitudes. The Fourier Transform converts a signal between time and frequency domains.

The Spectral Analysis Tools are Simcenter Amesim tools dedicated to frequency-domain analysis of signals generated by Amesim models.

FFT (Fast Fourier Transform)

Applying FFT to a signal transforms it into its frequency-domain representation, revealing the frequency components composing the signal. The FFT tool is applied directly to a 1D plot curve via the Analysis menu.

Spectral Map

A spectral map is a 2D map made of slices, each slice being the FFT of the source signal computed over a restricted, contiguous (possibly overlapping) time range ("moving FFT window"). Spectral maps are commonly used to analyze vibration effects on rotating machines during a velocity transient (run-up when speed increases, run-down when it decreases).

Like FFT, the Spectral Map facility is applied directly to a 1D curve via the Analysis menu of the Amesim plot (read the FFT section first). Selecting the menu item creates a cursor; clicking on the signal in the plot opens a setup dialog controlling signal segmentation (e.g. Number of FFTs, the number of slices).

PSD (Power Spectral Density)

PSD describes how a signal's power is distributed across its frequency components. Like FFT, it's applied directly to a 1D plot curve via the Analysis menu.

Order tracking

Order tracking is the natural analysis method for spectral maps: it extracts "order curves" from the spectral map into a 2D display, revealing potential system resonances.

Because the original signal is sampled at a fixed time step, order curves in the spectral map do not correspond to fixed-frequency spectral lines. Instead, order curves follow the relation X = order * Y (X and Y in compatible units, e.g. Hz and RPM) — e.g. the order-1 curve passes through (X=100 Hz, Y=6000 RPM), the order-2 curve through (X=100 Hz, Y=3000 RPM).

Fixed time sampling causes some loss of spectral energy: each point on an order curve corresponds to the frequency RMS of the FFT computed over a specified interval (whose length is a configurable parameter of the order-tracking computation).