Application of Vibration Analysis in Predictive Maintenance of Rotating Equipment

That Day, a Measurement Machine's Vibration Anomaly Sent Me from Heaven to Hell

I still remember several years ago, a critical measurement machine on our production line experienced a sudden drop in Cpk from 1.35 to 1.08, with DPMO directly jumping to 6210. The entire FAB was in chaos. After half a day of investigation, machine parameters hadn't changed, and the recipe hadn't been touched, yet no cause could be found. At that time, everyone was guessing if the machine was about to fail, feeling very unsettled. The production line supervisor was extremely worried, monitoring the line daily, fearing an unplanned shutdown at any moment. That kind of pressure, honestly, really makes you want to bury your head in the machine.

Where Was the Problem? That Unseen "Vibration"

To put it bluntly, many times, our equipment anomalies, especially in rotating equipment, actually all start from "vibration." You might think the machine is fine, but it might already be sending you distress signals. Just like when a person feels unwell, at first it might just be a minor ailment without feeling it, but if not addressed, it will gradually become a major problem.

This "shaking" we call vibration, and it is actually a barometer of the equipment's health. From motors, pumps, fans to bearings, anything that rotates—its vibration frequency and amplitude during rotation can tell you a lot.

So the key point is, we must learn to "understand" these vibrations. When Cpk suddenly drops, if you immediately check the machine for abnormal vibration, perhaps you can detect the problem early, instead of waiting until the machine fails to rush to the rescue.

How to Actually Do It? Master the "Stethoscope" Approach

When we perform Predictive Maintenance (PdM), vibration analysis is definitely your invaluable teammate. The simplest way is to regularly measure vibration data at several key points.

1. Baseline Establishment: First, when the machine is operating normally, measure its vibration data to establish a "health baseline." This is like your annual health check-up, knowing what your healthy indices are.
2. Regular Monitoring: Next, measure at regular intervals and compare it with your baseline. For example, if your motor's amplitude at the 1X RPM rotational frequency suddenly spikes from 1.5 mm/s to 4.0 mm/s, this indicates it might have an imbalance.
3. Spectrum Analysis: A more advanced approach is to look at the spectrum graph. The spectrum graph is like the machine's DNA; different failure modes will show anomalies at specific frequencies.

* 2X RPM Peak Increases: Could be misalignment, where the motor and pump are not aligned, leading to double the shaking during rotation.

* Multiple Peaks at Bearing Frequencies: The bearing might be about to fail, with damage to the internal rolling elements or inner/outer races.

So the key point is, through the changes in this data, you can determine if it's motor imbalance, misalignment, or a bad bearing, and then perform maintenance before a failure occurs.

The Most Common Pitfalls: Number Obsession and Human Negligence

One of the most common pitfalls I see is that people only look at the overall amplitude and not the spectrum. Last time, a new colleague of ours, seeing that the overall amplitude was not exceeding the limit, said the machine was fine. What was the result? The machine still failed! Later, a deep analysis revealed that although the overall amplitude hadn't changed, the bearing's fault frequency on the spectrum was already alarmingly high, simply because other frequencies were very low, which "averaged out" the overall amplitude.

Another big pitfall is "human negligence." Once, the vibration data of one of our pumps was consistently abnormal, but maintenance engineers changed the bearings several times to no avail. Later, when I personally went to the site to inspect, I discovered that after maintenance, the fastening screws weren't tightened, leading to a loose base, which generated additional vibration. This kind of frustrating situation, honestly, really leaves you dumbfounded.

So the key point is, data must be interpreted correctly, not just superficially; it must also be combined with on-site conditions to truly solve the problem.

One Thing You Can Do Today

Pick up your vibration meter and measure the vibration of the most critical machine on your production line!