How to Establish Vibration, Temperature and Oil Baselines for a Gearbox

A baseline is not a set of standard values. It is the set of readings this unit produces under conditions you have documented. Establish it while load and speed are under control, and attach three things to every record: the duty (load, speed, oil temperature), the measurement point and direction, and the instrument with its range. Leave out any one of the three and those numbers can never be compared again. From that moment on, every trend judgment holds only under the three-same rule: same duty, same measurement point, same instrument.

Why a generic alarm threshold cannot replace a baseline

A generic alarm threshold looks like a shortcut. It carries one defect that cannot be designed around: it knows nothing about your unit.

A generic vibration alarm line does not know which bearing your transducer sits on, which direction it faces, what load the gearbox was carrying, which instrument you used, or where the range was set. On one and the same machine, a reading taken on the bearing housing and a reading taken on a thin casing wall can differ by a wide margin, and a reading taken at part load does not match a reading taken at full load. Ignore all of that, compare against a single fixed figure, and the result errs in both directions: false alarms at light load, missed events at full load.

So the position taken here is straightforward: build this unit's own baseline first, and only then discuss when a reading deserves attention. Until that baseline exists, any threshold borrowed from elsewhere is somebody else's experience value, not your reference.

One boundary belongs in this section too. This article gives no alarm thresholds — not for vibration, temperature, sound pressure or oil condition — and it does not label any figure as "statutory" or "standard-required". A normative statement written without the verifiable text of the governing specification behind it is misleading, however confidently it is phrased.

Three dimensions, none of them optional

The three have to be read together. Vibration alone can miss degraded oil. Temperature alone can miss an early mesh fault. Oil alone can miss a loose mounting. Drop one dimension and the condition judgment is incomplete.

Measurement points and directions

Two placement rules keep readings comparable from one visit to the next:

1. Close to the bearing. Vibration energy leaves the machine through the bearing. The closer the transducer sits to it, the cleaner the signal. Mounting on a thin casing wall or on a cover plate folds extra structural response into the reading.
2. One reading in each of three directions: axial, horizontal, vertical. Different faults show up differently depending on direction, so a single direction leaves information out. How many measurement points to take is set by the installation manual for that model.

Every measurement point needs a physical mark. Fix position and direction on the machine itself with a scribe, a punch mark or a label, and take every later reading at that same spot, in that same direction. Without a mark, the second measurement is already not the same measurement point — this is the most frequent way a baseline fails.

When the baseline must be rebuilt

A baseline rests on one premise: the reference conditions stay unchanged. When any of the following four things happens, the old baseline is void from that point on:

1. Commissioning of a new unit — the first baseline is established here;
2. After an overhaul or replacement of a key component — new gears, bearings or shafts change the vibration and thermal behaviour of the system;
3. After an oil change — oil grade, viscosity and fill quantity can all change;
4. After a sensor is replaced or moved — the point or the instrument has changed, and the three-same rule is broken.

One more case is easy to overlook: a permanent change in operating condition, such as running long-term at a different load point after a process adjustment. Rebuild the baseline there too, or at minimum add a second baseline set for the new condition.

Conditions you must record alongside the readings

Readings without conditions are readings you cannot use. Every record carries these six items at least:

All six do the same job: they anchor the reading to a condition that can be reproduced.

Trend reading: look at the slope, not the single point

Once a baseline exists, there is exactly one way to judge: read the trend, not the individual reading.

A single high reading proves nothing on its own. It may be a temporary load peak, a fresh oil fill, or an instrument that has not warmed up. What deserves attention is this: under unchanged duty, several consecutive readings move in the same direction.

Two expressions cover it:

Δx = x_now - x_baseline

x_now is the current reading, x_baseline is the baseline value, and Δx is the deviation.

k = Δx / Δt

Δt is the interval since the baseline was set or since the previous record, and k is the rate of change. `k` is the basis for the judgment: a k near zero means stable; a k that stays positive while its magnitude grows means degradation is accelerating.

The shape of the trend carries more information than any number in it:

Note what this table delivers: a direction, not a threshold. How fast counts as fast, and how much rise counts as abnormal, depends on the model, the duty and the unit's own history. No universal figure answers it. This article gives a method for judging; the thresholds themselves have to come from this unit's own history.

Oil sampling: three points that decide whether the data is usable

The easiest place to get an oil baseline wrong is not the analysis. It is the sampling.

Three points:

1. A fixed sampling point. Draw from the same location every time — the same drain port or the same sampling valve. Oil drawn from different locations reflects wear at different components.
2. A fixed sampling moment. Cold machine or running machine, it has to be the same every time. Particle distribution in suspension differs with oil temperature.
3. A retained sample. Keep one sample from the same period for later re-checking or comparison.

One more: take the first sample after the new oil has run for a while, with the interval set by the lubrication article and the manufacturer's recommendation. A sample drawn immediately after an oil change mostly measures the new oil itself, not the condition of the machine.

Condition-by-condition judgment table

Field checklist

When establishing the baseline (all in one pass):

1. Confirm the unit is running steadily, with load and speed under control and recorded.
2. Mark the measurement points near the bearings and label all three directions.
3. Record instrument model, range and sensor serial number.
4. Take vibration (three directions), temperature (oil and bearing housing) and oil (sample plus retained sample).
5. Complete all three collections on the same day, so conditions are not mixed across records.

On routine inspection:

1. Confirm the point and direction match the baseline.
2. Record load, speed, oil temperature, ambient temperature and operating hours.
3. Use the same instrument on the same range.

When a trend change appears:

1. Calculate Δx and k, and establish whether this is a trend or a one-off fluctuation.
2. Check the duty records and rule out the effect of a load or temperature change.
3. When the specific fault type has to be identified, move to symptom diagnosis (link in the next section).

Common errors and where this article stops

Common errors:

1. Treating a generic threshold as a baseline and drawing a conclusion without a reference for this unit.
2. No physical mark on the measurement point, so every measurement is taken somewhere slightly different.
3. Readings recorded without the duty, leaving the data incomparable.
4. Continuing to use an old baseline after an overhaul or an oil change.
5. Comparing directly after an instrument change or a range change, with no bridging measurement.
6. Judging a fault from one high reading.
7. Sampling oil at an unfixed location or moment, which makes the data incomparable.

Where this article stops:

• The baseline definition, the three-same rule, point directions, the trend expressions and the record fields here apply to condition monitoring of industrial gearboxes.
• This article gives no vibration values, frequency values, alarm thresholds, temperature limits, oil-condition limits or sound-pressure thresholds.
• This article offers no universal answer to how fast counts as abnormal; a trend judgment has to draw on this unit's history and duty records.
• This article does not use unverified normative wording such as "statutory threshold", "standard-required" or "industry practice".
• This article does not replace the operating manual of the condition-monitoring instrument or the manufacturer's maintenance requirements.

Baseline record sheet

Fields 3 and 10 are the two most often skipped, and they are precisely the two that decide whether the next reading can be compared at all. Change the instrument range or move the sampling point and the earlier data is void.

About this article and the publisher

This article was written by the technical team of Chengfa Reducer (Changzhou) Co., Ltd. (CHINFAR) and published after the company's technical review. The H/B series industrial gearboxes discussed here are supplied by Chengfa Reducer (Changzhou) Co., Ltd.; CHINFAR products are produced and delivered on the joint manufacturing system shared with HUIGUANG.

Related products and technical material:

• H/B series industrial gearboxes
• Gearbox noise and vibration: working back from symptom to cause
• Gearbox lubrication guide: oil types, fill quantities and change intervals

Symptom diagnosis is handled in the second article and is not repeated here. Oil selection and fill quantities are handled in the third, also not repeated here.

Next step

Send the eleven fields above to sales@chinfar.com. If you do not have a baseline yet, three things come first: mark the measurement points, fix the instrument, and collect vibration, temperature and oil on the same day. Once those three are done, every later comparison has a reference to stand on.

To be explicit: condition judgment follows this unit's own history, this article provides no universal threshold, and it does not replace the operating manual of your condition-monitoring instrument.