A Whine, a Rattle, or a Hum: Which Points Where

Why this matters

Different fault families make recognizably different sounds, and sorting a noise into the right category in the first few seconds saves you from chasing the wrong mechanism. A whine is not a rattle is not a hum is not a knock, and each one has a short, specific list of likely causes. This tree sorts by the sound itself, heard directly by you, not by a customer's secondhand description of it (see the companion Reference on why that distinction matters).

Start here: describe it in one word, then confirm by listening yourself

Get close enough to hear the sound directly. Sort it into one bucket before you do anything else. If more than one bucket seems to apply, you likely have more than one thing wrong; note both and work them separately.

If it's a whine or a high-pitched squeal

A continuous, high-frequency tone.

• If it tracks with speed (rises and falls as the unit speeds up or slows down): suspect a bearing developing wear, or a belt slipping under load. Bearings start as a faint whine long before they progress to a rumble or a grind; catching it here means you have time to plan a replacement instead of reacting to a seizure.
• If it is constant regardless of speed: suspect an electrical source, a transformer or a winding humming at line frequency, or airflow noise from a restriction (a whistling gap, a partially blocked opening).
• If it only appears under load, not at idle: suspect a component being asked to work harder than it is built for right now, such as a motor nearing its limit or a pump fighting a restriction.

If it's a rattle or a clatter

An irregular, banging, changeable sound, not a smooth tone.

• If it comes and goes, or changes when you press on a panel or bracket: this is looseness, the single most common and cheapest cause to rule out. Check mounting hardware, panels, guards, and set screws before suspecting anything internal. See the companion article on vibration loosening a connection over time.
• If it is a single sharp clack once per revolution or once per cycle: suspect a discrete defect, a chipped part, a bent blade, or debris caught somewhere it should not be.
• If it is random and does not correlate with the machine's cycle at all: check for a loose object nearby (a panel, a piece of debris inside a housing, an item that fell into an enclosure) rather than assuming an internal fault.

If it's a hum or a buzz

A low, steady, continuous tone, often felt as much as heard.

• If it is localized to one specific spot (a single connection, a single component): suspect resistive heating at an electrical connection, or a mechanical part vibrating against something it touches. Localized heat paired with a hum at an electrical connection is a potential fire and shock hazard; de-energize and verify dead before you touch it.
• If it is general across the whole unit and has always been present at some level: this may simply be normal operating sound. Compare against a known-good matching unit before treating it as a fault at all; not every hum is a problem, though a hum that has clearly gotten louder over time than that same baseline is not something to wave off just because some hum is normal.
• If it only appears at one specific speed and clears above and below it: suspect resonance, not a failing part. See the companion Reference on telling resonance from a genuine mechanical fault before you replace anything.

If it's a knock or a thump

A single, discrete, repeating impact sound, heavier than a rattle.

• If it is synced exactly to the rotation or cycle (once per revolution, once per stroke): suspect a specific mechanical defect at that point in the rotation, such as a worn bearing race, a bent component, or a failing coupling.
• If it is irregular and unpredictable: suspect something loose that is free to move and impact on each cycle rather than a fixed, worn point.

Recap

1. Get close and confirm the sound category yourself; do not diagnose from a customer's description alone.
2. Whine: check speed-dependency first, then bearing wear vs. electrical/airflow source.
3. Rattle: rule out looseness first, it is the cheapest and most common cause.
4. Hum: check for localization (lead with safety if electrical), compare to a known-good baseline, and rule out resonance before condemning a part.
5. Knock: check whether it syncs to rotation (a fixed defect) or is irregular (something loose).

References

• ISO 13373-1, Condition monitoring and diagnostics of machines, Vibration condition monitoring, general procedures.
• NFPA 70B, Standard for Electrical Equipment Maintenance, on thermographic inspection of connections and terminals.
• See related: Reading a Vibration or Sound Signature as Its Own Diagnostic Method; The Difference Between a Resonance and a Genuine Mechanical Fault.