The FFT (Fast Fourier Transform) in Vibration Analysis

Definition: What is an FFT?

The Fast Fourier Transform (FFT) is a highly efficient mathematical algorithm used to transform a signal from the time domain into the frequency domain. In vibration analysis, this means converting a raw, complex time waveform (a plot of vibration amplitude versus time) into a frequency spectrum (a plot of vibration amplitude versus frequency). This transformation is the single most important and fundamental process in modern machinery diagnostics.

Why is the FFT Essential for Diagnostics?

A raw time waveform from a running machine is a complex jumble of many different vibrations all happening at once. It is nearly impossible to look at this signal and determine the health of the machine. The FFT acts like a prism, separating this complex signal into its individual frequency components. The result is a clear, actionable chart that allows an analyst to see:

• What frequencies are present?
• How much energy (amplitude) is at each frequency?
• What is the relationship between these frequencies?

Because different mechanical and electrical faults (like unbalance, misalignment, bearing defects, and looseness) each generate vibration at very specific, predictable frequencies, the FFT spectrum provides a direct roadmap to the root cause of a problem.

Key Parameters of an FFT Analysis

To acquire a useful FFT spectrum, a vibration analyst must define several key parameters on the data collector or software:

1. Fmax (Maximum Frequency)

This is the highest frequency that will be included in the spectrum. It must be set high enough to capture the highest frequency fault you are looking for (e.g., high-frequency gear mesh or bearing tones).

2. Resolution (Lines of Resolution)

This determines the level of detail in the spectrum. It is defined as the number of discrete frequency “bins” or data points that will be calculated across the Fmax. A higher number of lines (e.g., 3200 or 6400) results in better frequency resolution, meaning the ability to separate two vibration frequencies that are very close together. High resolution is critical for diagnosing beat frequencies or analyzing sidebands in gearbox analysis.

3. Averaging

Because machine vibration can fluctuate, a single “snapshot” FFT can be misleading. Averaging involves acquiring multiple FFTs in quick succession and then averaging them together. This process reduces random noise and provides a much more stable, repeatable, and representative spectrum of the machine’s true condition.

4. Windowing

A window function (like Hanning) is a mathematical weighting applied to the time waveform data before the FFT is calculated. It is a signal processing technique used to minimize an error called “spectral leakage,” ensuring that the amplitude and frequency of the signals in the spectrum are as accurate as possible.

Interpreting an FFT Spectrum

A trained analyst interprets an FFT spectrum by looking for characteristic patterns:

• A large peak at 1x the running speed indicates unbalance.
• A large peak at 2x the running speed often points to misalignment.
• A series of harmonics (peaks at 1x, 2x, 3x, 4x, etc.) is a classic sign of mechanical looseness.
• A high-frequency peak with sidebands spaced at the running speed is a tell-tale sign of a gearbox or bearing fault.
• A raised “floor” of broadband noise can indicate cavitation in a pump or friction.

By comparing the current FFT spectrum to a baseline taken when the machine was healthy, analysts can easily spot changes and diagnose developing problems long before they become critical failures.

← Back to Main Index