Vibration Analysis Basics for Maintenance

Vibration analysis is the practice of measuring how a rotating machine shakes and reading that signal to judge its health. Every fault a machine can develop, imbalance, misalignment, looseness, a failing bearing, changes how it vibrates in a recognizable way. Catch the change early and you can plan the repair instead of reacting to a breakdown. It is the single most informative condition-monitoring technique for rotating equipment.

This matters because rotating machines rarely fail without warning; they fail after weeks or months of a vibration signature getting worse while everything else looks normal. A pump does not care that its bearing is on borrowed time until the day it seizes. Vibration is the signal that spans that gap. This guide is the entry-level primer: what vibration tells you, the three ways it is measured, how the readings are collected on a route, and how standards decide whether a number is fine or a problem. The deeper diagnostic work, reading the frequency spectrum, builds on these basics.

What does vibration tell you about a machine?

Vibration tells you both that something is wrong and, with more analysis, what is wrong. A healthy machine has a low, steady vibration baseline. As a fault develops, the level rises and the pattern changes, and because different faults excite different motions, the change carries a fingerprint. Imbalance shows up as motion once per revolution. Misalignment adds motion twice per revolution and in the axial direction. Looseness rattles across many multiples of running speed. A worn bearing generates its own high-frequency tones. You do not need the full diagnosis to act, a rising overall level alone tells you to look closer, but the signal contains the diagnosis when you are ready to read it.

The reason vibration is such a rich signal is that it is mechanical cause and effect made measurable. A force inside the machine, an unbalanced rotor, a rubbing part, a spalled bearing race, produces motion the machine's structure transmits to its surface, where a sensor picks it up. Nothing has to fail for the signal to appear; the force exists as soon as the defect does. That is what makes vibration an early-warning tool rather than a failure alarm, and it is why vibration sits at the center of most condition-based maintenance programs.

Acceleration, velocity, or displacement, which do you measure?

You measure all three in principle, but you pick the one that best matches the frequency of the fault you care about, because the same vibration looks big or small depending on which parameter you use. The three are mathematically linked, velocity is the rate of change of displacement, acceleration the rate of change of velocity, so an instrument can convert between them, but each one emphasizes a different part of the frequency range.

For everyday machine health on general rotating equipment, velocity in mm/s is the workhorse, which is why the severity standards are written in it. Acceleration is the parameter for catching early bearing and gear faults, because those live at high frequency where acceleration is most sensitive, the reason most modern sensors are accelerometers, with the instrument integrating to velocity when needed. Displacement matters most on large, slow machines with fluid-film bearings, where the quantity of interest is literally how far the shaft moves. Choosing the parameter is really choosing the frequency band you want to see clearly. The sensor that captures each is covered in vibration sensor types.

How is vibration data collected?

Most vibration data is collected one of two ways: by a technician walking a route with a portable analyzer, or by permanently mounted sensors streaming data continuously. Route-based collection is where most programs start. A technician carries a data collector and an accelerometer to a defined list of machines, takes a reading at fixed points and directions on each, and uploads the data to software that stores and trends it. Continuous monitoring puts fixed sensors on critical or hard-to-reach machines and streams readings automatically, which removes the labor and catches fast-developing faults between route visits.

Either way, the discipline that makes the data useful is repeatability. A reading is only comparable to last month's if it was taken at the same point, in the same direction, with the same mounting. Move the sensor an inch, change from a magnet to a probe, or read a different axis, and the number changes for reasons that have nothing to do with the machine. This is why serious programs fix permanent measurement points and standardize mounting, a topic large enough to have its own guide in accelerometer mounting methods since the mount also sets the highest frequency you can trust.

What is overall vibration versus a spectrum?

Overall vibration is a single number summarizing how much a machine is shaking; a spectrum breaks that same vibration into the individual frequencies inside it. The overall level, often an RMS velocity in mm/s, is the simplest measurement and the backbone of a screening program. It answers one question well: is this machine shaking more than it should, or more than it used to? Trend it over time and a rising overall level flags a machine that needs attention, without any deep analysis.

What the overall level cannot tell you is why. Two very different faults can produce the same overall number. To separate them you need the spectrum, which uses a mathematical transform to show how much vibration sits at each frequency, so imbalance at running speed, misalignment at twice running speed, and a bearing tone all appear as separate peaks. The basic program trends overall levels to find machines in trouble; the diagnostic step reads the spectrum to name the fault. That deeper step is its own subject, covered in vibration spectrum analysis.

How do standards decide what level is a problem?

Standards decide by defining vibration limits and sorting a machine into zones from good to dangerous. The international framework is the ISO 20816 series, which measures and evaluates machine vibration and places a machine in one of four zones: Zone A for newly commissioned machines, Zone B acceptable for long-term operation, Zone C not suitable for continuous long-term running and needing remedial action, and Zone D at levels that can cause damage. The limits depend on machine size, speed, and mounting, and they are written in RMS velocity for general industrial machines. ISO 20816 replaced the older ISO 10816 and ISO 7919 standards, folding both into one series.

Standards give you an absolute yardstick, but your own trend is often the sharper tool. A machine can sit inside an acceptable zone and still be degrading if its level has doubled over a few months. The strongest programs use both: the standard to set alarm thresholds and to compare against similar machines, and the machine's own baseline to catch a rising trend before it crosses any absolute limit. For the detail on the severity bands, see the ISO 10816 vibration standards and how they carry into ISO 20816.

How do you start a vibration program?

You start small and repeatable, then grow. A first program does not need continuous sensors on every machine; it needs a defined route, consistent readings, and someone who acts on the trend.

1. Pick the machines that matter. Start with critical and hard-to-replace rotating equipment where a failure hurts most. Do not try to cover everything at once.
2. Define measurement points. Set fixed points and directions on each machine, typically radial-horizontal, radial-vertical, and axial at each bearing, and mark them so every reading comes from the same spot.
3. Choose the parameter and sensor. Velocity in mm/s for overall health, acceleration for bearing detail. An accelerometer with the instrument integrating to velocity covers both.
4. Baseline every point. Record readings on known-good machines. The baseline is what every later reading and alarm is judged against.
5. Collect on a schedule. Walk the route at a fixed interval with consistent mounting and technique. Consistency is what makes the trend trustworthy.
6. Set alarms from standards and trend. Use ISO zones for absolute limits and a rise over baseline for early warning; alarm on both.
7. Act and close the loop. When a point alarms, investigate, read the spectrum, check alignment and balance, and tie the finding to a work order so the reading turns into a repair.

Where vibration analysis fits your reliability program

Vibration analysis is the backbone of condition monitoring for rotating equipment because it gives the earliest, richest read on machine health and scales from a clipboard route to a plant-wide streaming system. A basic overall-level program is cheap to start and catches a large share of developing failures; the diagnostic depth is there when you want it. It is the first predictive technology most maintenance teams should build, and it pays back in avoided failures and planned repairs.

The limit is rarely the readings, it is connecting them to everything else the plant knows about the machine. A vibration trend in a standalone tool cannot see the work-order history, the run hours, or the production schedule that give it context. Pulling vibration together with the plant's other data into one operational view is where machine monitoring platforms like Harmony fit, so a rising trend lands next to the maintenance record and the schedule rather than in a silo. It layers onto the systems you already run, with no rip-and-replace. For the broader strategy, see predictive maintenance or read the platform overview.