Check the Causes of Gear Grinding Cracks Before You Scrap the Batch

Before any sectioning, photograph the flank under low magnification and mark the orientation of every crack. The pattern tells you more than a hardness number. A single crack running across the whole tooth height, roughly perpendicular to the grinding direction, points at a thermal event during one pass. Cracks that sit only on one flank of a few adjacent teeth usually mean a localized problem: a hard spot in the case, a wheel that loaded up, or a coolant nozzle that lost its aim.

Count the cracked teeth and note whether they are neighbors or scattered. Neighboring teeth share one wheel pass and one coolant path, so a cluster narrows the search to that zone. Scattered single cracks on many teeth suggest a batch condition instead: case depth variation, retained austenite above target, or a wheel that is too hard for the material.

Measure the crack direction against the grinding lay. Cracks perpendicular to the lay are typical of tensile stress from a fast thermal cycle. Cracks that run along the lay often start at a pre-existing defect, such as a grinding mark from a previous operation, a forging lap or a heat treatment quench crack that only opened up after stock removal.

One practical note. If you see the same pattern on gears from two different heat treatment batches, stop looking at the furnace. The common factor is then the grinder setup, the wheel specification or the coolant, and you can check those in a single shift.

Sectioning is the only way to separate a heat treatment problem from a grinding problem. Cut through the worst crack, mount it, polish to 1 μm and etch in 2–4% nital. Look at three things in one pass: case depth on both flanks, the microstructure near the surface, and the shape of the crack tip.

A crack tip that follows prior austenite grain boundaries and shows a dark, over-tempered band along its side is a grinding burn crack. A crack that runs through a coarse martensitic case with no tempering band nearby is more likely a heat treatment quench crack, reopened by grinding. The two need completely different fixes, and hardness alone will not tell them apart.

Check case depth on the left and right flank at the same height. Uneven effective case depth is common on gears that were ground after a long heat treat cycle, and the thinner flank cracks first. On a 5 mm module gear, a 0.2 mm difference between flanks is enough to shift the failure to one side.

Look at the retained austenite level under the ground surface. High austenite softens the case, the wheel cuts deeper than expected and the specific energy rises. That chain produces burn without any obvious change in the grinding program.

If the crack depth stays above the finish stock allowance, the defect was made before grinding. If it reaches below the final flank surface, the grinding operation either opened it or created it. That single depth measurement decides which department owns the problem.

Grinding burn has three practical levels, and each one looks different under the microscope. Reading them correctly stops you from re-tempering perfectly good parts or, worse, shipping cracked ones.

The mild form is temper burn. The surface is slightly darker after etching at low magnification, and hardness drops by 1–2 HRC in a shallow band. It rarely cracks on its own, but it reduces the fatigue limit of the flank and it signals that the heat balance is already marginal.

The medium form is re-hardening burn. The surface layer re-austenitizes and quenches in the same pass, so you get a hard, brittle white layer with fresh untempered martensite. Hardness may read higher than the drawing. Cracks appear at the boundary between the white layer and the softer material below it. This type is the most common cause of the fine, closely spaced cracks seen on production gears.

The severe form is a full burn with visible discoloration before etching. It comes with deep cracks, sometimes several tenths of a millimeter, and often with a change in part geometry. Parts at this level are usually scrap.

Check the white layer thickness, not just its presence. A layer under about 5 μm can sometimes be removed by a controlled re-grinding pass with a softer wheel and lower depth of cut. Anything thicker than that is a stress raiser that will not survive service.

Most grinding cracks that reach us for review are not caused by a bad furnace. They come from specific energy that is too high at the contact zone. The fix is usually a combination of a softer wheel, a faster dressing cycle and more coolant delivered exactly where the arc of contact forms.

Dressing is the parameter people skip. A dull wheel rubs instead of cutting, and rubbing converts nearly all the spindle power into heat in a thin layer. Shorten the dressing interval and check the dress depth. On a vitrified CBN wheel, a light continuous dress often holds the flank temperature far more stable than a heavy intermittent one.

Coolant has to reach the contact zone, not the top of the tooth. Aim the nozzle at the point where the wheel leaves the flank and keep the velocity high enough to break the vapor film. Flow rate alone is a poor guide. A bucket test at the nozzle shows the real delivery after the lines, filters and rotary unions take their share.

Depth of cut and feed are the last knobs. Splitting one heavy pass into two lighter passes lowers the peak temperature sharply, at the cost of cycle time. For gears that already show temper burn, that trade is usually worth taking before you change the heat treatment cycle.

Check the wheel specification against the material hardness. A wheel that is one grade too hard will produce burn on a 60 HRC case even with perfect coolant. One grade softer often removes the burn entirely and adds only a small amount of wheel wear.

When the grinding side is clean, go back to the furnace records. Temper twice at a lower secondary temperature, control the quench speed, and check that the case depth target matches the final part, not the pre-grind blank. Extra stock means extra grinding time and extra heat.

Carburizing potential is a frequent culprit. A boost stage that runs too rich leaves a high-carbon surface layer with coarse carbides and high retained austenite. That layer grinds hot and cracks easily. Carbon profile data from a test coupon is more useful here than a hardness traverse.

Quench agitation and fixturing decide distortion, and distortion decides how much stock the grinder has to remove on one side. A gear that is 0.15 mm out of round will be ground unevenly, with one flank taking a deeper cut and running hotter.

Sub-zero treatment after quench reduces retained austenite and stabilizes the case. It is not needed for every gear, but on thin-rim or high-speed parts it often removes the crack risk that no grinding change can fix.

Document the sequence. When a crack appears after a supplier change, a furnace rebuild or a new wheel batch, the record of what changed is the fastest route to the cause.

After you change a parameter, verify on a small number of gears and inspect them properly. Nital etch plus a 10× visual check on every tooth is the fastest field test. It shows temper burn and re-hardening burn without sectioning.

For a stronger verification, cut one tooth from the first, middle and last part of the trial run. Compare case depth and white layer thickness across the run. A stable result over three parts is more meaningful than a perfect single sample.

Keep a control sample from the cracked batch. If the new settings look good, run the control through the same nital etch and compare. This removes the doubt that the etchant or the lighting changed the result.

Track hardness on the finished flank. A drop of more than 1 HRC against the pre-grind value means the grinding cycle is removing tempered material, and the burn risk is still present.

Run a short fatigue or load test when the gear is safety-critical. Metallography shows the layer; a load test shows whether the remaining stress state survives the service loads you actually expect.