The same gear can be called grade 6, grade 7 or AGMA 10 depending on the standard. How the three systems map, what each grade actually controls, and what every step up costs.
Photo: 先従隗始 / CC0 / Wikimedia Commons
A drawing that says “accuracy grade 6” is not a specification. It is the name of a table in a standard, and three standards are in daily use in export programmes: GB/T 10095.1 in China, DIN 3962 in Germany and much of Europe, and AGMA 2000-A88 or its successor ANSI/AGMA 2015 in North America. The numbers do not run in the same direction, they do not use the same reference diameter, and a tolerance that looks generous in one system can be tight in another.
This matters more than it sounds. A supplier quoting to GB/T 7 and a buyer expecting DIN 7 are usually talking about the same part; a buyer expecting AGMA 10 and receiving GB/T 7 may be receiving something noticeably coarser than intended. Misreading the grade does not show up in the price, only in the noise measurement at the end of the line, which is the most expensive place to discover it.
Each standard defines a set of individual deviations and then a grade that bounds them for a given size. The four that decide how a gear sounds and how long it lasts are: single pitch deviation, which measures tooth-to-tooth spacing error and drives high-frequency noise; total profile deviation, which covers form error on the flank; total lead deviation, which describes how the tooth is twisted along its width and therefore how the load distributes; and runout, which turns once per revolution and produces the low-frequency component of gear noise.
In GB/T 10095.1 and DIN 3962 the tolerances are tabulated by grade and by diameter, with grade 0 the finest and grade 12 the coarsest. AGMA runs the opposite way: a higher number is a finer gear, with AGMA 15 at the top of normal commercial practice. That inversion alone causes a remarkable share of the confusion in international enquiries, because “grade 10” means a precision gear in AGMA and a rough one in DIN.
The mapping below is the one to keep on the desk. It is approximate by nature — the standards define different quantities over different reference lengths, so exact equivalence is impossible — but it is accurate enough to prevent a procurement error, and it is what experienced gear buyers use in practice.
| GB/T 10095.1 | DIN 3962 | AGMA 2000-A88 | Typical process to reach it |
|---|---|---|---|
| 4 | 4 | 13–14 | ground and superfinished, temperature-controlled shop |
| 5 | 5 | 12–13 | ground teeth, form-ground or generating grinding |
| 6 | 6 | 11–12 | ground teeth, stable heat treatment |
| 7 | 7 | 10–11 | hobbed or shaped, then hardened and corrected |
| 8 | 8 | 9–10 | hobbed, heat treated, no grinding |
| 9 | 9 | 8–9 | hobbed, general commercial quality |
| 10 | 10 | 7–8 | rough hobbing, non-critical drives |
Two practical notes. First, the AGMA column shifts by up to one grade depending on tooth size and on whether the older 2000-A88 or the newer 2015 system is being used, so always confirm which edition a North American customer is quoting. Second, Chinese suppliers often state a GB/T grade for the tooth flank but say nothing about runout or lead, which are frequently the deviations that actually limit the assembly.
Accuracy is bought in hours of grinding and in rejection rate, not in a catalogue surcharge. Moving from GB/T 8 to 7 means adding a hard-finishing operation, usually grinding or honing after heat treatment, and that operation is where capacity disappears. Moving from 7 to 6 means controlling distortion tightly enough that the grinding allowance is uniform, which in turn means fixtures, quench control and a stress-relief step. Moving from 6 to 5 starts to require thermal stability in the shop and a metrology lab that can actually measure what is being promised.
| Grade reached | Added operation | Cost index | Scrap risk |
|---|---|---|---|
| GB/T 8, hobbed only | — | 100 | low |
| GB/T 7, ground after heat treat | grinding | 135–150 | medium |
| GB/T 6, controlled distortion + ground | fixtured grinding | 170–195 | medium-high |
| GB/T 5, ground + superfinish | honing, stable shop | 230–280 | high |
| GB/T 4, precision ground | temperature control, 100% inspection | 320–400 | very high |
The numbers are indices, not prices, but the shape is reliable: each grade roughly adds a third to the previous one at the top of the range. This is why specifying grade 5 everywhere is a common and expensive mistake. A gear that only transmits torque at 1,500 rpm in a tractor gearbox does not benefit from a tolerance designed for a 14,000 rpm e-axle; the money would be better spent on case depth control and on the root fillet.
Where accuracy grade really bites is noise. At 2,000 rpm a small pitch error produces a low-frequency modulation that is masked by engine noise. At 14,000 rpm the same error generates a tone at the mesh frequency and its sidebands, right where the ear is most sensitive, and no amount of sound insulation fixes a transmission error generated at the tooth.
So specify by function rather than by habit. For an e-axle input gear running above 10,000 rpm, GB/T 6 or better with a stated lead modification and a documented transmission-error target is defensible. For a commercial-vehicle gearbox at 2,000–2,600 rpm, GB/T 7 with controlled lead crowning is usually sufficient and far cheaper. For agricultural and off-highway transmissions, GB/T 8 with attention to case depth and root geometry outperforms a needlessly tight grade on a poorly supported shaft.
Four items remove almost all ambiguity, and a supplier who cannot answer them is not ready for an export programme: the standard and edition, not just the grade number; the grade for profile, lead, pitch and runout listed separately, since many drawings specify only one; the reference diameter and the inspection method, whether that is double-flank roll, single-flank or coordinate measurement; and the acceptance criterion for noise, usually a mesh-frequency amplitude limit on a stated test rig at a stated speed and load.
Add one more line that costs nothing and prevents most disputes: require the first-article inspection report to list measured values against each tolerance, not a tick box. A report that says “conforms” tells the buyer nothing about how much margin is left, and margin is what predicts whether the tenth batch will still conform.
