Spiral bevel gears are widely used in aeroengine, helicopter, and high-power industrial transmissions because their curved teeth provide gradual engagement, high contact ratio, and favorable load sharing. The root fillet remains a fatigue-critical region: a local tensile stress concentration can initiate a crack, reduce transmission accuracy, and eventually cause tooth fracture. For the same gear geometry and operating point, however, different standards can produce substantially different root bending stresses and safety factors.
Three calculation systems are especially relevant: the Chinese aviation industry standard HB/Z 89.1–89.4, issued in 1985 and derived from ISO practice for aeroengine bevel gears; ISO 10300-1:2014 and ISO 10300-3:2014; and ANSI/AGMA 2003-B97. They share the same underlying cantilever-beam idealization but differ in critical-section construction, load position, reference plane, correction factors, and allowable-stress format. These differences explain why published comparisons have sometimes ranked ISO and AGMA conservatism in opposite orders.
Evaluation Criteria and Mechanical Idealization
Two equivalent forms of root-strength verification are used. The first compares the calculated root stress with an allowable value, \(\sigma_F \le \sigma_{FP}\). The second compares a calculated safety factor with a minimum value, \(S_F \ge S_{F\min}\). HB and ISO use an explicit minimum safety factor, although the numerical values and their placement in the allowable-stress equation differ. AGMA does not use \(S_{F\min}\) in the same way; it introduces a reliability factor \(Y_Z\), with values of 0.70, 0.85, 1.00, 1.25, and 1.50 corresponding approximately to failure probabilities of 50%, 10%, 1%, 0.1%, and 0.01%.
HB and ISO convert the bevel gear to a virtual cylindrical gear at the midspan, use the 30° tangent construction to locate the critical root section, and apply the nominal load at the tooth tip. AGMA uses outer-end bevel-gear dimensions, determines the critical section through an inscribed parabola, and places the load at the critical load point rather than at the tip. The parabolic construction and the different load position change the chord tooth thickness and bending moment arm, so the resulting tooth-form and stress-correction factors are not interchangeable between the standards.
| Item | HB | ISO 10300 | AGMA 2003-B97 |
|---|---|---|---|
| Reference gear | Virtual cylindrical gear at midspan | Virtual cylindrical gear at midspan | Outer-end bevel gear |
| Critical section | 30° tangent | 30° tangent | Inscribed parabola |
| Load position | Tooth tip | Tooth tip | Critical load point |
| Geometry representation | Several separate geometry factors | Several separate geometry factors | Mainly one geometry factor \(Y_J\) |
| Reliability measure | Minimum safety factor \(S_{F\min}\) | Minimum safety factor \(S_{F\min}\) | Reliability factor \(Y_Z\) |
Stress Formulations and Correction-Factor Structure
In HB, the root stress and allowable stress are written as
$$
\sigma_F=\sigma_{F0}K_AK_vK_{F\beta},\qquad
\sigma_{F0}=Y_{Fa}Y_{sa}Y_\varepsilon Y_\gamma Y_iY_{LC}\frac{F_{mt}}{b m_{mn}}Y_b
$$
$$
\sigma_{FP}=Y_\theta Y_X Y_{ST}\frac{\sigma_{F\lim}}{S_{F\min}}
$$
where \(F_{mt}\) is the tangential force on the virtual cylindrical gear, \(b\) is the face width, \(m_{mn}\) is the normal module at midspan, and \(\sigma_{F\lim}\) is the bending fatigue limit. ISO uses a similar basic-stress structure but adds an inter-tooth load-sharing factor:
$$
\sigma_F=\sigma_{F0}K_AK_vK_{F\beta}K_{F\alpha},\qquad
\sigma_{F0}=Y_{Fa}Y_{sa}Y_\varepsilon Y_{BS}Y_{LS}\frac{F_{mt}}{b m_{mn}}
$$
$$
\sigma_{FP}=\sigma_{F\lim}Y_{ST}Y_{NT}Y_{\delta relT}Y_{RrelT}Y_X
$$
AGMA expresses the stress directly in terms of torque and outer-end dimensions:
$$
\sigma_F=\frac{2000\,T_1K_AK_vY_XK_{F\beta}}{b\,d_{e1}m_{et}Y_\beta Y_J}
$$
$$
\sigma_{FP}=\frac{\sigma_{F\lim}Y_{NT}}{S_FK_\theta Y_Z}
$$
Here \(T_1\) is the input torque, \(d_{e1}\) is the outer pitch diameter of the virtual cylindrical gear, \(m_{et}\) is the outer transverse module, and the factor 2000 is a unit-conversion term. The important structural difference is that AGMA concentrates tooth form, load sharing, effective face width, inertia, and contact-line geometry into \(Y_J\), whereas HB and ISO retain separate factors for these effects.
Load-Related Factors
The application factor \(K_A\) accounts for external overload. HB tabulates only uniform and light-vibration driver classes, whereas ISO and AGMA use the same wider four-level table. The numerical values are identical where the classes overlap, but ISO and AGMA extend the range to moderate and heavy vibration.
| Driver / driven machine | HB | ISO | AGMA |
|---|---|---|---|
| Uniform / uniform | 1.00 | 1.00 | 1.00 |
| Uniform / light vibration | 1.25 | 1.25 | 1.25 |
| Light / light vibration | 1.35 | 1.35 | 1.35 |
| Moderate / moderate vibration | Not tabulated | 1.75 | 1.75 |
| Heavy / heavy vibration | Not tabulated | 2.25 or higher | 2.25 or higher |
The dynamic factor \(K_v\) is a more important source of divergence. HB and AGMA both use empirical curves indexed by accuracy grade, but the grade numbering runs in opposite directions: in HB, a lower \(Q_H\) denotes higher precision, whereas in AGMA a higher \(Q_V\) denotes higher precision. ISO uses a dimensionless reference speed \(N\), divides operation into four resonance regions, and includes effective pitch deviation, mesh stiffness, and reduced mass. As a result, the same nominal gear can fall into different dynamic regions under the three methods.
For the helix load-distribution factor \(K_{F\beta}\), HB and AGMA use the same mounting-based table when the contact-pattern shift is unknown. For full load, both simply supported gears give 1.00, one simply supported gear gives 1.00, and neither simply supported gives 1.00; for light load the corresponding values are 1.05, 1.10, and 1.25; for the test gear at theoretical full load they are 1.20, 1.32, and 1.50. ISO instead writes
$$
K_{F\beta}=\frac{K_{H\beta}}{K_{F0}},\qquad
K_{F0}=0.211\left(\frac{r_c}{R_m}\right)^{0.279}\log_{10}\!\left(\sin\beta_m\right)+0.789
$$
where \(r_c\) is the cutter radius, \(R_m\) is the midspan cone distance, and \(\beta_m\) is the spiral angle. ISO therefore couples the bending factor to the contact-stress factor \(K_{H\beta}\), the load pattern, and the longitudinal curvature. This is why an ISO calculation can produce a much larger \(K_{F\beta}\) than HB or AGMA even when the mounting description is the same.
ISO also introduces the inter-tooth load-sharing factor \(K_{F\alpha}\), which HB and AGMA do not use explicitly. It depends on the virtual total contact ratio \(\varepsilon_{v\gamma}\), pitch deviation \(f_{pt}\), running-in allowance \(y_a\), and mesh stiffness \(c_\gamma\):
$$
K_{F\alpha}=
\begin{cases}
\displaystyle \frac{\varepsilon_{v\gamma}}{2}\left[0.9+0.4\frac{c_\gamma(f_{pt}-y_a)}{K_AK_vK_{H\beta}/b_v}\right], & \varepsilon_{v\gamma}\le 2,\\[8pt]
\displaystyle 0.9+0.4\sqrt{\frac{2(\varepsilon_{v\gamma}-1)}{\varepsilon_{v\gamma}}\frac{c_\gamma(f_{pt}-y_a)}{K_AK_vK_{H\beta}/b_v}}, & \varepsilon_{v\gamma}>2.
\end{cases}
$$
Tooth Geometry, Root Section, and Size Effects
HB carries seven geometry-related factors: tooth-form factor \(Y_{Fa}\), stress-correction factor \(Y_{sa}\), load-point factor \(Y_\varepsilon\), inertia factor \(Y_i\), load-sharing factor \(Y_\gamma\), longitudinal-curvature factor \(Y_{LC}\), and root-stress longitudinal-distribution factor \(Y_b\). With \(S_{Fn}\) the chord thickness at the critical section and \(h_{Fa}\) the bending moment arm to the tip load, the tooth-form factor is
$$
Y_{Fa}=\frac{6(h_{Fa}/m_{mn})\cos\alpha_{Fan}}{(S_{Fn}/m_{mn})^2\cos\alpha_n}
$$
and the HB stress-correction factor is
$$
Y_{sa}=H+\left(\frac{S_{Fn}}{h_{Fa}}\right)^{0.4}\left(\frac{S_{Fn}}{\rho_f}\right)^{0.2}
$$
where \(\rho_f\) is the fillet radius at the critical point and \(H=0.18\) for a 20° pressure angle. The longitudinal-curvature factor is the reciprocal of the ISO curvature term, \(Y_{LC}=1/K_{F0}\). The factor \(Y_b\) converts the average cantilever stress into the maximum root stress, while \(Y_\gamma\) describes how load is shared between teeth when more than one pair is in contact.
ISO replaces \(Y_\gamma\) with \(Y_{LS}\), a load-sharing factor based on contact-line lengths and peak loads in the tip, middle, and root contact ellipses. It adds \(Y_{BS}\) for non-uniform stress along the face width. The ISO overlap factor \(Y_\varepsilon\) uses the same symbol as the HB load-point factor but has a different role: it transfers the tip load to a specified reference point and is capped at 0.625 when the virtual overlap ratio exceeds unity.
$$
Y_\varepsilon=
\begin{cases}
0.25+0.75/\varepsilon_{v\alpha}\ge 0.625, & \varepsilon_{v\beta}=0,\\
0.25+\dfrac{0.75}{\varepsilon_{v\alpha}}-\varepsilon_{v\beta}\left(\dfrac{0.75}{\varepsilon_{v\alpha}}-0.375\right)\ge 0.625, & 0<\varepsilon_{v\beta}\le 1,\\
0.625, & \varepsilon_{v\beta}>1.
\end{cases}
$$
AGMA removes most of these separate terms. Its geometry factor \(Y_J\) contains the tooth-form term, effective face width, load-sharing ratio, inertia effect, and the radii to the load and pitch points. The longitudinal-curvature factor \(Y_\beta\) is numerically the same as the ISO curvature term and is therefore the reciprocal of the HB \(Y_{LC}\). This condensation makes AGMA the simplest of the three methods to apply, but it also makes the individual physical effects less transparent.
The size factor also behaves differently. HB sets \(Y_X=1\) for outer module \(m_t\le 1.588\) mm and uses \(Y_X=1.225m_t^{-0.25}\) above that, so \(Y_X\) decreases with module. ISO uses material-dependent linear expressions in \(m_{mn}\). AGMA uses a separate size-factor curve that increases with outer module, opposite to the HB trend. In the worked example below, the HB size factor is 0.6883 and the ISO value is 0.9975, a difference that directly affects the allowable stress.
Material, Surface, and Service-Life Factors
ISO is the only one of the three methods that explicitly separates material notch sensitivity and root-surface condition. The relative notch-sensitivity factor \(Y_{\delta relT}\) compares the actual gear material with the standard test gear, and the relative surface-condition factor \(Y_{RrelT}\) depends on root roughness \(R_Z\) and material. For hardened steel at \(R_Z=20\ \mu\mathrm{m}\), the ISO surface factor is about 0.9567; in the worked example \(Y_{\delta relT}=1.0081\). HB and AGMA do not include these two factors explicitly.
Service-life treatment is also asymmetric. HB uses a temperature factor \(Y_\theta\): it is 1 below 121 °C and \(394/(273+\theta_M)\) above that, where \(\theta_M\) is the peak gear-blank temperature. ISO uses a life factor \(Y_{NT}\) read from material-specific fatigue curves. AGMA uses a single piecewise expression:
$$
Y_{NT}=
\begin{cases}
2.75, & N_L \lt 10^3,\\
6.1514\,N_L^{-0.1192}, & 10^3\le N_L \lt 3\times10^6,\\
1.683\,N_L^{-0.0323}, & N_L\ge 3\times10^6,
\end{cases}
$$
where \(N_L\) is the equivalent number of stress cycles. AGMA also uses a temperature factor \(K_\theta\), which is the reciprocal of the HB \(Y_\theta\). Because one appears in the numerator and the other in the denominator of the allowable-stress expression, their physical effect is consistent: higher temperature reduces allowable bending stress.
Worked Example: Stress and Safety-Factor Ranking
Consider a spiral bevel gear pair with \(z_1=27\), \(z_2=37\), outer transverse module \(m_t=7.96\) mm, midspan spiral angle \(\beta_m=35^\circ\), normal pressure angle \(\alpha_n=20^\circ\), face width \(b=62\) mm, and design speed \(n=8500\) r/min. The input power is 5 kW, the equivalent stress-cycle number is \(N_L=10^9\), the bending fatigue limit is \(\sigma_{F\lim}=650\) MPa, the tooth-flank roughness is 4.8 \(\mu\)m, the root roughness is 20 \(\mu\)m, the oil temperature is 120 °C, both gears have the same hardness, and both are simply supported.
| Quantity | HB | ISO | AGMA |
|---|---|---|---|
| \(K_A\) | 1.000 | 1.000 | 1.000 |
| \(K_v\) | 1.000 | 1.8867 | 1.2578 |
| \(K_{F\beta}\) | 1.000 | 1.4519 | 1.0215 |
| \(K_{F\alpha}\) | — | 1.000 | — |
| \(Y_{Fa}\) | 2.3285 | 2.3285 | — |
| \(Y_{sa}\) | 1.6531 | 1.8300 | — |
| \(Y_\varepsilon\) | 0.5694 | 0.6250 | — |
| \(Y_J\) | — | — | 0.2450 |
| \(Y_\beta\) | — | — | 1.0448 |
| \(Y_{ST}\) | 1.775 | 2.000 | — |
| Calculated \(\sigma_F\) / MPa | 524.67 | 1301.8 | 365.17 |
| Computed \(S_F\) | 1.3314 | 0.9607 | 1.3351 |
The ISO stress is about 2.48 times the HB value and 3.56 times the AGMA value. The dominant contributors are the dynamic factor and the helix load-distribution factor: ISO gives \(K_v=1.8867\) and \(K_{F\beta}=1.4519\), compared with 1.000 and 1.000 in HB and 1.2578 and 1.0215 in AGMA. ISO also gives a slightly larger stress-correction factor, 1.830 versus 1.6531, and a larger overlap factor, 0.625 versus 0.5694. AGMA remains lowest because \(Y_J\) replaces several separate geometry terms and because its dynamic and load-distribution factors are lower in this case.
The safety-factor ranking is not the same as the stress ranking. ISO produces the highest calculated stress but also the highest allowable-stress basis; HB applies its own minimum safety factor and size factor; AGMA uses \(Y_Z\) rather than \(S_{F\min}\). Consequently, the ISO safety factor is 0.9607, below unity, while HB gives 1.3314 and AGMA gives 1.3351. A higher nominal stress therefore does not by itself imply a lower safety factor unless the allowable-stress definitions are placed on a common basis.
The same conclusion holds when tooth count and module are varied. Table 4 compares seven gear sets with \(z_1/z_2\) from 39/53 down to 21/29 and \(m_t\) from 5.513 mm to 10 mm. All other conditions remain as in the worked example.
| Set | \(z_1/z_2\) | \(m_t\) / mm | HB \(S_F\) | ISO \(S_F\) | AGMA \(S_F\) |
|---|---|---|---|---|---|
| 1 | 39/53 | 5.513 | 1.3489 | 0.5929 | 1.1580 |
| 2 | 36/49 | 5.972 | 1.3504 | 0.6415 | 1.2009 |
| 3 | 31/42 | 6.935 | 1.3139 | 0.7509 | 1.2946 |
| 4 | 27/37 | 7.960 | 1.3314 | 0.9607 | 1.3351 |
| 5 | 26/35 | 8.270 | 1.2998 | 1.1039 | 1.3425 |
| 6 | 25/34 | 8.600 | 1.2800 | 1.2522 | 1.3745 |
| 7 | 21/29 | 10.000 | 1.1517 | 1.5978 | 1.4603 |
For many teeth and a small module, HB gives the highest safety factor; ISO gives the lowest. For few teeth and a large module, ISO gives the highest safety factor and HB the lowest. ISO is also the most sensitive to simultaneous changes in tooth count and module, rising from 0.5929 to 1.5978 across the seven sets. AGMA rises more moderately, from 1.1580 to 1.4603, while HB falls from 1.3489 to 1.1517. The statement that one standard is always more conservative is therefore not supported by these calculations.
Parametric Sensitivities
Changing a basic design parameter does not move the three standards in the same direction. Table 5 summarizes the calculated-stress trends over the examined ranges.
| Parameter increased | HB \(\sigma_F\) | ISO \(\sigma_F\) | AGMA \(\sigma_F\) |
|---|---|---|---|
| Face width \(b\) | Decreases | Decreases | Increases |
| Spiral angle \(\beta_m\) | Non-monotonic | Non-monotonic | Increases |
| Pressure angle \(\alpha_n\) | Increases | Increases | Decreases |
| Pinion tooth count \(z_1\) | Decreases | Decreases | Decreases |
| Module \(m_t\) | Decreases | Decreases | Decreases |
| Design speed \(n\) | Decreases | Decreases | Decreases |
The face-width reversal is notable. HB and ISO contain an explicit \(1/b\) term in the nominal stress, and their geometry factors do not overcome it in the examined range, so a wider face reduces calculated stress. In AGMA, the explicit \(1/b\) dependence is offset by \(K_{F\beta}\) and by the geometry factor \(Y_J\); over the studied range, the net effect is an increase in calculated stress with face width. A designer therefore cannot assume that increasing face width always lowers root stress without checking the standard-specific geometry and load-distribution terms.
The pressure-angle reversal has a similar origin. In HB and ISO, increasing \(\alpha_n\) reduces the total contact ratio and changes the load-point factor in a way that raises calculated stress. In AGMA, the geometry factor \(Y_J\) changes in the opposite direction over the same range, so the stress falls. The spiral angle is the most sensitive parameter of all because it strongly affects overlap ratio, load sharing, and—in ISO—the resonance region used to evaluate \(K_v\). HB and ISO show a non-monotonic response with an extremum, whereas AGMA increases monotonically over 0° to 40°.
The allowable-stress side is also standard-dependent. HB allowable stress depends on tooth count, module, speed, face width, spiral angle, and pressure angle. ISO allowable stress is independent of speed but depends on material, surface condition, and life. AGMA allowable stress depends mainly on tooth count when the two gears have equal hardness; for equal hardness, tooth count no longer changes the allowable stress. These differences explain why stress and safety-factor trends can diverge even when the same nominal geometry is used.
Design Implications and Limits of the Comparison
AGMA has the most compact calculation structure: \(Y_J\) absorbs several geometric effects that HB and ISO treat separately, and the number of explicit correction factors is the smallest. This makes AGMA convenient for general commercial design and for preliminary sizing. HB and ISO retain more separate factors, including load sharing, longitudinal stress distribution, notch sensitivity, and root-surface condition, and are therefore better matched to high-load aeroengine applications where the additional physical resolution is useful.
The choice between HB and ISO should not be based on a blanket conservatism ranking. In the present calculations, HB is more favorable for many teeth and small module, whereas ISO and AGMA are more favorable for few teeth and large module. ISO is the most sensitive to tooth count and module because its dynamic factor can switch resonance regions and its overlap and load-sharing factors change strongly with geometry. HB is comparatively stable in the high-tooth-count range but loses safety margin as tooth count falls and module rises.
For spiral bevel gears in aero transmissions, ISO and HB are appropriate where explicit load sharing, material sensitivity, and surface condition must be represented. HB is particularly relevant to designs with many teeth and small module, while ISO offers a more detailed treatment of dynamic resonance and load distribution. AGMA is best suited to general industrial and commercial practice, where a compact calculation and a reliability-factor format are preferred.
Two cautions follow. First, safety factors from different standards must not be compared directly unless the allowable-stress basis, failure probability, and minimum-reliability provision are normalized. Second, the numerical comparisons here are applications of the three standards to the same nominal gear sets; they are not strain-gauge measurements or finite-element validation. The rankings are therefore conditional on the stated geometry, material, roughness, load, and speed ranges. Within those ranges, the dominant conclusion is robust: the three standards describe the same physical failure mode but not the same numerical stress, and the difference is large enough to change a design decision.