Ask what limits the life of a harmonic drive and the answers you get are fatigue, wear and ratcheting. Those are the failure modes and we have written them up. But in ordinary service, inside rating, the thing that degrades first is usually the grease — and the wear that eventually shows up in the tooth flanks is a consequence of that rather than an independent process.
So the obvious question is how often to change it. Look for the number and you will mostly not find one.
Why the number is missing
Not evasion, for once. A relubrication interval is not a property of the drive, and a manufacturer who prints one is quietly assuming a duty cycle on your behalf. Grease life depends on how fast the input turns, how hard the mesh is loaded, how hot it gets and what it is exposed to — and those vary more between two installations of the same part number than they do between part numbers.
The honest position is that the interval is a property of your application. We publish nothing on it either, which is worth saying plainly rather than leaving you to notice.
The four things that actually set it
Input revolutions, not operating hours. The wave generator turns at the full input speed — a hundred times output speed at 101:1. A slow, high-torque axis running eight hours a day may be turning its bearing fewer times than a fast one running two. Count revolutions.
Temperature, which dominates everything else. Grease degrades by oxidation, and oxidation rate roughly doubles for every 10 to 15 °C. The widely used rule of thumb is that every such step halves grease life. A drive running at 70 °C is not slightly worse off than one at 40 °C; it is a small number of halvings worse off. This is why the operating temperature range on a datasheet is a life specification and not just a survival limit — ours runs −20 to 80 °C on the AS-SJ1 servo joints and −10 to 60 °C on the AS-SF force-sensing series, and those two numbers imply different maintenance regimes.
Load, by the same cube-ish weighting that governs the bearing. The cube-weighted average torque you calculated for sizing is the right input here too. A duty cycle with hard transients works the grease harder than its time-average suggests.
Orientation and sealing. A vertically mounted drive lets grease migrate away from the mesh in a way a horizontal one does not. An unsealed installation in a machining environment ingests coolant mist and swarf, and contaminated grease is worse than no grease because it becomes a lapping compound.
Who owns the decision depends on what you bought
This is worth checking before it becomes a surprise in service. Across our strain wave gear range the three variants are not equivalent on this point:
- AS-GS, component set. Wave generator, flexspline and circular spline. You are assembling it into your own housing, so the grease selection, the fill quantity and the interval are yours to specify. This is the most flexible option and the one that hands you a responsibility.
- AS-GB, with output bearing. Same, plus the output bearing. Same responsibility.
- AS-GU, gear unit. Enclosed and greased as supplied. The decision has been made for you, which is usually what you want unless you have a reason to override it.
The signature to watch for
Grease degradation announces itself before it does damage, and the signal is one you can already measure. No-load running torque rises. It is the same figure we discussed in why the efficiency number is not what you will get, and it is the cheapest condition-monitoring channel a joint has: command a slow rotation with the load removed, log the current, and trend it.
A drive whose running torque has climbed noticeably above its commissioning value is telling you the grease is going, weeks or months before noise, backlash growth or debris appear. If you record one number at installation, record that one.
What to ask, and what we can tell you
- Which grease, by specification rather than brand. Base oil type, thickener, temperature range. Brand names get discontinued; specifications can be matched.
- What fill quantity, for this frame. A volume or a mass, not “pack the cavity”.
- What was the reference condition? If a supplier does quote an interval, it is against some assumed speed, load and temperature. Get those three, then scale to yours.
We would rather give you the variables than a number that pretends to know your machine. Tell us the duty cycle, the ambient and the mounting orientation and we will work the interval through with you against a specific part number — and say so if the answer is that the enclosed AS-GU variant is the one you should be buying.
Related: How harmonic drives fail · Where precision dies at the mounting flange.