Gears and Bearing Seats: Heat-Treatment Distortion Is the Real Enemy of Precision

Article

June 15, 2009

The tolerances on a gear drawing are finished-state tolerances: profile, lead, bore cylindricity, face runout. What decides whether they can be met is how much material is left to grind after heat treatment — and distortion is invisible before the furnace.

Gear Accuracy Is Accuracy After Heat Treatment

Tolerances on gears and bearing seats are specified in the finished state: profile tolerance, lead tolerance, bore cylindricity, face runout. What actually decides whether those can be held is how much allowance survives heat treatment for grinding. During carburising and quenching, a part undergoes structural transformation and thermal stress at the same time: thin walls grow, thick sections shrink, asymmetric parts bend. None of it is visible before the furnace.

A high tolerance bore is not produced by a better grinder; it is produced by a process that knew how far the bore would move before it was ground. Process design for gears is therefore distortion prediction plus allowance allocation. Experienced shops build a distortion database by material, wall-thickness ratio and loading pattern, then compensate the pre-heat-treatment dimensions in reverse. Shops without that data can only leave more allowance and take extra grinding passes. Cost rises, and too much allowance grinds through the carburised case and drops the hardness.

Three Sources of Distortion: Material, Geometry, Loading

The first is material. Two heats of the same 20CrMnTi grade can differ in hardenability band width enough to change distortion by thirty percent. Sound practice is to require a hardenability band report from the mill and select toward the lower-middle of the band.

The second is geometry: uneven wall thickness, an offset keyway, asymmetric holes in the web all produce uneven cooling during quenching. These should be raised at the drawing stage — design for manufacturing review can often convert an asymmetric feature into a symmetric one or add a process rib. The third is loading: the fixture determines the quenchant flow field, cooling rate differs between basket centre and edge, and parts in the same load can distort in opposite directions.

Grinding Allowance Is Calculated, Not Guessed

Too little allowance and distortion cannot be cleaned up, leaving black skin after grinding. Too much and grinding time and cost climb while the carburised case risks being cut through. Correct allowance equals expected distortion plus stock removal plus a safety factor, and the expected distortion has to come from measurement, not from a handbook.

The practical route is a process validation run: load ten parts in one cycle, measure the size distribution before and after, derive mean shift and standard deviation, then set the allowance. In production, sample by batch and plot the distribution on a control chart; a shift in the mean means the furnace or the material has changed. None of this is new, but few shops do it properly — which is exactly the part of a custom machining service that is hardest to replace.

Batch Consistency: One Good Part Proves Nothing

The characteristic failure on gears is not a single out-of-tolerance part but batch dispersion. A part measures at mid-tolerance, yet across a batch one end drifts out — usually the sum of thermal growth, tool wear and fixture wear. After four hours of continuous machining, spindle thermal growth can reach tens of microns, enough to consume the whole tolerance band.

Control has three layers: first-article confirmation, in-process sampling, last-article verification; plus thermal warm-up before production and tool-life management by cutting length rather than by breakage. For bearing seats held to IT6, measurement uncertainty itself must be counted: a two-degree difference between gauge and workpiece is enough to read a one-micron error on a fifty-millimetre bore.

Gear and Bearing Fit: Two Parts, One Problem

The fit between a gear shaft and its bearing is decided jointly: shaft tolerance, bearing inner-ring tolerance, and the roundness of both. The common error is watching only the shaft dimension and ignoring roundness and surface roughness. At the same interference, a shaft with poor roundness deforms the inner ring into an ellipse, bearing clearance changes, and noise and temperature rise follow.

Parts like these are therefore better machined by one supplier, or at least within one datum system. Split between two shops, it is very common for each to pass inspection and the assembly to fail. A direct question works well in supplier evaluation: are the gear and the bearing seat machined in the same clamping on the same machine?

Inspection: A CMM Cannot Measure Tooth Form

Gear accuracy cannot be carried by a CMM alone. A CMM measures bore size, face runout and keyway position, but profile and lead tolerances require continuous probing along the involute, which means a gear measuring centre or a dedicated involute tester. Plenty of quotations promise full CMM inspection when the tooth form has never actually been verified.

That is a familiar source of acceptance disputes: the part passes on CMM data and still whines when it goes into the housing. The cause may be profile crowning, a helix angle deviation or accumulated pitch error, and none of those sit in a standard CMM routine.

A workable acceptance plan is layered. The first article gets a full report from the gear measuring centre covering profile, lead, pitch and radial runout. Production runs use double-flank composite inspection as a fast screen, with periodic sampling back at the measuring centre. Double-flank machines are cheap and quick enough for one hundred percent inspection, but they only show a composite error and cannot say which element is out, so the two methods have to work together.

One caution about the measurement itself: a gear measuring centre is more sensitive to calibration and temperature than a CMM, and twenty degrees plus or minus one is the baseline. Looking at the metrology room's temperature log during an audit tells you more than the machine's nameplate.

Four Questions for a Buyer

First, is carburising and quenching done in-house or outsourced, and if outsourced does the heat treater hold CQI-9? Second, can measured distortion data be provided, with mean and standard deviation grouped by material and wall thickness? Third, how is grinding allowance set, and has a validation run been done? Fourth, how is batch consistency assured — is there SPC and tool-life management?

Few suppliers answer all four cleanly, but those that do usually quote realistically, because they know they will not lose the money on scrap. When shortlisting a China precision machining factory or a Chinese CNC parts manufacturer, use heat-treatment capability as the first filter; it tells you far more than an equipment list.