Ideally a vibration signal will be free of any noise or interference that would hinder analysis. Unfortunately, this is never the case in the real world; there are always constant and uncontrollable changes to the vibration in a machine that add noise to the signal. Figure 1, below, shows a noise-free vibration signal that can be produced by a rotating machine with a mass imbalance. Figure 2 shows the same signal, but this time it has noise and is a more accurate representation of the data that would be collected.
Figure 2. Illustration of a typical vibration signal, showing noise and interference
Data analysis would be much simpler if we did not have to deal with noise and interference, however there are methods to tackle this issue. Averaging is a good method for minimizing the effects of noise, for isolating frequencies of interest and for revealing specific patterns in the data. Several averaging techniques are available in most vibration analysis software, including:
- Linear averaging
- Peak-hold averaging
- Time synchronous averaging
Linear averaging is done by dividing the vibration signal into specific lengths of time. Analysis is performed on each segment of the signal and their spectral values are averaged. The averaging will typically minimize any noise in the signal, depending on the number of averages taken.
Peak-hold averaging is not a true form of averaging. This method keeps the peak or highest value at each frequency as data is being collected. It is a good method for detecting or monitoring for a sudden increase in amplitude at a particular frequency; with linear averaging the increase in amplitude may be averaged out of the data.
Time synchronous averaging differs from linear averaging because the waveform is averaged instead of the spectral data. This method requires the use of a speed encoder or tachometer signal so that the waveform data can be synchronized with the machine speed. Any signals that are not a integer multiple of the machine speed are averaged to zero. This method is useful for detecting synchronous frequencies in the vibration data.
More complex forms of averaging are also available, depending on the vibration analysis software being utilized. Each averaging technique has its own set of advantages and disadvantages; it is important to recognize this in order to implement the correct averaging method, if needed.