The overlooked “collection time” — why low-speed faults get missed

How should vibration data collection time be determined?

• Too short a duration may fail to fully capture the equipment’s operational characteristics;

• Too long, however, does not necessarily yield better results.

The conclusion is: for equipment operating at different speeds, the vibration waveform collection time is typically not the same and it directly affects the accuracy of subsequent fault analysis.

This is not an arbitrary setting, but rather determined by a combination of rotational speed, frequency resolution requirements and fault mechanism, as outlined below.

Ensuring sufficient revolutions are captured

First, we need to consider a fundamental operating parameter — rotational speed. Equipment running at different speeds experience vastly different numbers of revolution cycles over the same collection time.

• For high-speed equipment, multiple revolution cycles can be completed within a very short time;

• For low-speed or even ultra-low-speed equipment, however, the same collection time may not even cover a single full revolution.

  
  

For low-speed equipment, to guarantee analysis accuracy, we need to collect data covering at least 4–6 revolutions.

As can be seen:

• High-speed equipment rotates quickly, capturing many revolutions in a very short time.

• Low-speed or ultra-low-speed equipment rotates very slowly. If we still collect data for only 0.4 seconds, we may not even capture a single revolution, making many periodic faults (such as gear faults and bearing faults) impossible to identify.

Therefore:

Low-speed equipment requires longer collection time to ensure a sufficient number of revolutions, thereby guaranteeing the accuracy of data analysis.

Improving frequency resolution at low frequencies

FFT frequency resolution Δf=1/T, where T means collection time. For Example:

• 0.5 seconds yields a resolution of 2 Hz,

• 2 seconds yields 0.5 Hz,

• 4 seconds yields 0.25 Hz.

The longer the collection time, the higher the frequency resolution.

Low-speed equipment requires a longer collection time to achieve higher frequency resolution, enabling closely spaced fault frequencies to be clearly distinguished and improving diagnostic accuracy.

This is particularly critical because many key diagnostic frequencies are concentrated in the low-frequency range.

• These include rotational frequency components (1X and 2X), which are commonly used to diagnose mechanical faults such as imbalance, misalignment.

• As well as bearing characteristic frequencies such as the Fundamental Train Frequency (FTF) and Ball Pass Frequency of the Outer race (BPFO) and so on.

If the collection time is too short, the insufficient frequency resolution causes:

• These frequencies to overlap in the spectrum.

• Difficult to distinguish between the rotational frequency and its harmonics.

• The effectiveness of envelope spectrum analysis is also significantly compromised.

Take a piece of equipment operating at approximately 1500 rpm as an example:

According to basic vibration theory, when the sampling rate remains unchanged, increasing the waveform collection length means longer collection time, thereby improving frequency resolution.

With the sampling rate held constant, envelope spectrum analysis using a waveform acquired with 64k samples — benefiting from higher frequency resolution — can more clearly identify the FTF.

In contrast, when using 16k samples, the lower frequency resolution makes the FTF peak more susceptible to interference from adjacent frequency components and spectral leakage, resulting in reduced fault feature identification capability and an increased risk of missing early-stage faults.

Therefore:

low‑speed equipment requires longer collection time to achieve adequate frequency resolution, ensuring that fault characteristics can be accurately identified.

Ensuring complete capture of impact events

On low‑speed shafts of gearboxes, many typical faults — such as gear pitting, uneven loading, bearing spalling, eccentricity, and looseness — often generate impacts that occur only once per revolution.

Take a low‑speed shaft running at 120 rpm as an example:

One full revolution takes 500 ms. If the waveform collection time is only 500 ms, a single collection may capture just one impact, making it difficult to determine any repetitive pattern. However, if the collection time is extended to 5 seconds, up to 10 impacts can be recorded, and the recurring interval between them becomes clearly visible — greatly facilitating accurate fault pattern recognition.

Therefore:

for low-speed equipment, sufficient collection time is essential to ensure complete capture of fault-related impact events and avoid missing critical diagnostic information.

High‑speed Equipment focuses on high‑frequency information instead of longer time

For high‑speed equipment, monitoring priorities typically include gear mesh frequency (GMF), high‑frequency bearing fault signatures, envelope analysis, and high‑frequency impacts. These characteristic frequencies are significantly higher than the equipment’s rotational frequency.

Take a high‑speed shaft running at 3000 rpm as an example:

Its rotational frequency (1X) is 50 Hz, while the gear mesh frequency (GMF) can be as high as 1500 Hz, and bearing fault‑related frequencies are typically distributed across the 500–5000 Hz band.

To effectively capture these high‑frequency components, the key requirements are

• a sufficiently high sampling rate;

• a sufficiently wide frequency band;

• rather than an excessively long collection time.

Therefore:

for high‑speed equipment, a relatively high sampling rate combined with a shorter collection time is generally sufficient to meet analysis requirements.

Summary

In summary, the vibration data collection time should not be fixed, but rather flexibly adjusted based on equipment speed, fault mechanisms, and analysis requirements.

• Low‑speed equipment requires longer time to ensure sufficient revolutions and frequency resolution, thereby accurately identifying low‑frequency fault characteristics.

• High‑speed equipment, on the other hand, needs high sampling rates to capture high‑frequency information, with time that can be appropriately shortened.

Only by configuring collection parameters according to equipment characteristics can a reliable data foundation be established for subsequent fault diagnosis.