Spiral bevel gears are used where smooth torque transfer and high load capacity are required, but their curved, tapered teeth make the final contact pattern sensitive to both flank geometry and the subsurface condition produced by carburizing. In a production batch of 12CrNi3A steel spiral bevel gears, the grinding-in test produced a contact pattern positioned too low along the tooth height. The pattern did not meet the requirement that the contact area cover at least 40% of the tooth height, and the gear generated mild abnormal noise and vibration. The teeth showed no magnetic-particle indications, the chemistry was within specification, and the final profile and lead data met grade 7. The failure therefore raised a more specific question: had final grinding removed the carburized case from the surfaces that actually carry load?
Gear Design, Duty, and Manufacturing Route
The gear had 41 teeth, a midpoint normal module of 5 mm, a midpoint spiral angle of 37°, left-hand spiral direction, a large-end transverse module of 7.349 mm, and accuracy grade 7. Its material was 12CrNi3A, a low-carbon Ni-Cr carburizing steel. The specified surface hardness was 60–63 HRC, core hardness 30–40 HRC, and effective case depth 1.0–1.4 mm. The process route was: forged blank → rough turning → normalizing → quenching and tempering → semi-finish turning → tooth milling → carburizing and quenching → shot blasting → finish turning → tooth grinding → grinding-in → journal grinding → final inspection.
Tooth milling left a 0.30 mm grinding allowance on the common normal per flank. After carburizing, the parts were high-temperature tempered, quenched in fast bright quenching oil, and cryogenically treated at −100 °C to promote martensitic transformation in the high-Ni steel. A 25 mm diameter furnace coupon from the same cycle had 61.0–62.5 HRC surface hardness and 1.560 mm effective case depth, meeting the process target. That coupon result confirmed the furnace cycle could produce an adequate case on a cylindrical specimen, but it did not reproduce the local stock removal on a curved, tapered bevel tooth.
Inspection Results
Material and baseline quality. Two tooth-block samples, A and B, were cut from the gear and divided into small-end, mid-face, and large-end pieces. Macroscopic examination and fluorescent magnetic particle testing found no tooth cracks. The measured chemical composition met GB/T 3077—2015 for 12CrNi3 steel.
| Element | C | S | Si | Mn | P | Cr | Ni |
|---|---|---|---|---|---|---|---|
| Measured | 0.16 | 0.004 | 0.24 | 0.35 | 0.007 | 0.82 | 2.93 |
| GB/T 3077—2015 range for 12CrNi3 | 0.10–0.17 | ≤0.025 | 0.17–0.37 | 0.30–0.60 | ≤0.025 | 0.60–0.90 | 2.75–3.15 |
Nonmetallic inclusion ratings were A 0.5, B 0, C 0, and D 0.5 according to GB/T 10561—2023, all within acceptable limits. The inclusions were mainly small sulfides and globular oxides. The failure was therefore not associated with a gross material or cleanliness defect.
Hardness at the tip versus hardness in the case. Hardness measured at the tooth tip with a portable Leeb tester, converted to HRC, was 60.2–62.5 HRC across the small-end, mid-face, and large-end zones, within the 60–63 HRC requirement. That result is a convenient shop-floor check, but it does not sample the loaded pitch and root surfaces. At 0.1 mm below the pitch surface, the converted microhardness was below 60.0 HRC, and sample A on both flanks and sample B on the left flank fell below 58.0 HRC. The near-surface hardness had therefore been reduced even where the tooth tip still passed.
Effective case depth. Effective case depth was measured perpendicular to the surface under a 9.8 N load, with 550 HV1 as the boundary according to GB/T 9450—2025. At the pitch, values ranged from 0.655 to 1.047 mm; only the mid-face right flank of sample B reached the 1.0 mm lower limit. Root values were 0.115–0.333 mm and groove values 0.259–0.478 mm. GB/T 38805—2020 requires the root case depth to be at least two-thirds of the pitch value; for a 1.0 mm pitch minimum this gives about 0.67 mm. Every root measurement was far below that limit, and the shallowest root case was only 0.115 mm.
| Sample | Axial position | Left flank | Groove | Right flank | ||
|---|---|---|---|---|---|---|
| Pitch | Root | Pitch | Root | |||
| A | Small end | 0.757 | 0.115 | 0.290 | 0.659 | 0.226 |
| Mid-face | 0.967 | 0.202 | 0.294 | 0.830 | 0.205 | |
| Large end | 0.655 | 0.118 | 0.259 | 0.565 | 0.227 | |
| B | Small end | 0.983 | 0.333 | 0.478 | 0.810 | 0.236 |
| Mid-face | 0.738 | 0.153 | 0.309 | 1.047 | 0.255 | |
| Large end | 0.880 | 0.318 | 0.416 | 0.924 | 0.310 | |
Microstructure. At the mid-face of sample A, the pitch region consisted mainly of acicular and lath martensite with small granular carbides and retained austenite. The core was lath martensite and bainite with small blocky undissolved ferrite at grain boundaries. Under JB/T 6141.3—1992, the pitch martensite plus retained austenite was rated grade 2, carbide grade 1, and core microstructure grade 3. The undissolved ferrite indicates that the austenitizing hold was not sufficient to dissolve all ferrite, so that hold should be extended. At the root arc, a visible etching transition at about 0.20 mm corresponded to the measured case-depth boundary.
Profile and lead accuracy. Profile and lead were measured on a Z6 gear measuring center on three teeth spaced 120° apart. Each flank was sampled at 5 points along the profile and 9 points along the face width. The maximum profile/lead deviations were −21.5, −19.6, and −20.6 μm, within the grade 7 drawing requirement of GB/T 11365—2019. Final grinding therefore produced an acceptable pointwise geometry, but this did not prove that the design flank curvature was correct.
Failure Mechanism
The primary cause was excessive stock removal in final tooth grinding. The process reserved 0.30 mm on the common normal, but the actual removal at the tooth groove approached 1.0 mm, more than three times the allowance. The grinding-wheel profile and the need to generate the specified flank curve increased removal at the root and groove. As a result, the carburized case at the pitch was reduced below 1.0 mm at most positions, and at the root and groove it was almost completely removed. The furnace coupon had 1.560 mm case depth and passed, but the loss occurred after carburizing, during grinding to final geometry.
Case depth controls more than a hardness number. The gradient from the surface to the core determines contact fatigue resistance at the pitch and bending fatigue resistance at the root. A pitch surface with less than 60 HRC at 0.1 mm, and some flanks below 58 HRC, is more prone to pitting and wear. At the root, the case was only 0.115–0.333 mm, so the hardened case that supports bending fatigue was largely absent in the region of highest bending stress. The low contact pattern concentrated load toward the root and reduced the effective contact area; together with the hardness variation, this produced the abnormal noise and vibration during grinding-in.
Two factors overlapped. The first was grinding-induced case loss, which weakened the loaded flanks and root. The second was an error in the tooth profile parameters at the simulation design stage, which shifted the contact pattern even though the measured profile and lead deviations were within grade 7. Tolerance compliance is pointwise; it does not guarantee the conjugate flank relationship that positions the contact pattern. The low pattern was therefore not solely a heat-treatment or grinding defect, but the case loss explains the hardness drop and the vibration risk.
Process Control and Corrective Actions
- Increase the case-depth margin before grinding. Extend both the boost and diffusion stages by 1–2 h, targeting a pitch effective case depth of at least 1.70 mm after carburizing and quenching. This leaves enough margin for final grinding while keeping 1.0–1.4 mm at the pitch and at least 0.67 mm at the root after grinding. Verify the result on sectioned parts at the small end, mid-face, and large end, not only on a coupon.
- Control actual stock removal. Match the reserved common-normal allowance to measured distortion, optimize the grinding-wheel profile and grinding path at the root and groove, and monitor stock removal by zone. If the wheel shape requires nearly 1.0 mm at the groove, either the case-depth margin or the process route must change.
- Extend the austenitizing hold. Eliminate blocky undissolved ferrite in the core and improve core uniformity before quenching.
- Recheck the design-stage tooth profile parameters. Use a contact-pattern test in addition to CMM profile/lead acceptance. Grade 7 data alone can miss a design curvature error that shifts the contact pattern.
- Change the shop-floor acceptance check. Tooth-tip hardness should not be the only hardness check. Add case-depth mapping and near-surface hardness at the pitch and root, because tip hardness can pass while the working case is lost.
Conclusion
The failure of the 12CrNi3A spiral bevel gear was driven mainly by excessive grinding stock removal. The 0.30 mm allowance was exceeded, especially at the root and groove, reducing the effective case depth to 0.115–0.333 mm at the root and 0.655–1.047 mm at the pitch; near-surface hardness fell below 60 HRC and in places below 58 HRC. The resulting hardness gradient was inadequate for the contact and bending loads, and the low contact pattern plus a design-stage profile parameter error produced noise and vibration. Material chemistry, cleanliness, and final profile/lead accuracy were acceptable. A larger pre-grind case-depth margin, tighter control of stock removal, elimination of undissolved ferrite, and correction of the design flank parameters are the appropriate countermeasures.