You may have heard of using a Bode plot to determine natural frequencies and possible resonance conditions in complex machinery. The idea can be somewhat intimidating until the basics are better understood. In order to arrive at such a plot, measurements have to be set up with an instrument with a vibration pickup and a speed (phase) reference sensor.
Begins an article by Dennis H. Shreve on the ReliabilityWeb blog. Shreve continues:
Once instrumentation is in place to capture data, the machine is either started up or coasted down from its operating speed. A coast down measurement is usually taken in the field, as the conditions are a little better controlled. For this measurement, the machine is taken to its normal operating speed, power is cut, and it is allowed to coast until it stops. Depending on how long the process takes from running speed to stop, the measurement intervals can either be set on a RPM delta or time interval. The idea is to get enough data to be able to pick out natural frequencies that show up in the process.
Keep in mind the relationship between lines of resolution, averages and Fmax. The higher the Fmax, the faster data collection will be. However, higher Fmax values will result in less resolution on separating frequency. The more averages and the higher lines of resolution will result in more time for measurement. As an example, if we set up a data capture with an Fmax of 1 KHz (60,000 CPM) and 4 averages and 800 lines, we will see a data capture time of 2 seconds. With this setup, we will be able to see a separating frequency of 225 RPM. The deciding factor on parameters for the measurement depends on the coast down time of the machine. If it is coasting down faster than 225 RPM in 2 seconds, we might need to reduce lines of resolution or averages or increase Fmax in order to be able to gather more points of data during the coast down. As you can see, it is somewhat of an iterative process, and it will become easier with additional experience.
Below is a good example of coastdown measurement that shows captured FFT data. This is in a special waterfall display with a cursor trend to show velocity amplitude versus machine speed. The vibration measurement setup was for a one average 400-line FFT with Fmax of 60,000 CPM. Data capture time was quite fast at 0.4 seconds per sample. The machine was taken to its normal running speed at 2374 RPM before power was cut. It was then allowed to coast to a stop while data were collected on amplitude, speed, and phase. Note the prominent rise in amplitude at approximately 600 RPM. (Data were captured at speed intervals of approximately 60 RPM in this setup.)
This might be a good starting point for determining a natural frequency and resonance condition, but the Bode plot is a much better analysis tool. The Bode plot with its dual display of amplitude and phase against speed is a positive indicator of resonance, as the rise in vibration amplitude will be accompanied by a 180-degree phase shift across the “skirts” of the peak, with a 90-degree change right at the peak. See the same data below with the aid of the Bode display.
Here we can pinpoint the actual natural frequency peak with much more accuracy. It is determined to be 577 RPM, and the phase plot confirms an accompanying 180-degree shift across the frequency.
For a variable speed machine, this information would tell us not to operate it at 577 RPM for any length of time, as vibration forces can be amplified by 30 times at this running speed and will most likely result in serious damage to components.