Design Basics: Shaft Sizing, Fatigue, and Critical Speed

Diameter is a result, not a starting assumption

A shaft’s diameter is often the most visible design choice, but it is not governed by torque alone. In a transmission, gears and pulleys apply transverse forces that bend the shaft, while couplings, brakes, and driven equipment transmit torque. Bearings constrain the shaft at discrete locations; shoulders, keyways, retaining-ring grooves, and changes in diameter disturb the local stress field. Speed adds another set of questions: how much will the shaft deflect, and how close is the operating range to a lateral critical speed?

These effects interact. Increasing diameter usually reduces nominal stress and deflection, but it can change bearing selection, gear geometry, mass, inertia, and packaging. A defensible design therefore begins with the load path and layout, then proceeds through strength, fatigue, deformation, and dynamic checks. A preliminary diameter is a useful checkpoint—not a final answer.

For enclosed gear drives, ANSI/AGMA 6001-F19, Design and Selection of Components for Enclosed Gear Drives, is a relevant industry reference. Its scope makes it useful when assessing components in that application. It does not replace analysis of the particular shaft’s geometry, loads, material, or operating conditions, and designers should consult the published standard for its exact requirements and applicability.

Build the load picture before calculating stress

Start with a free-body diagram. Mark bearings, gears, pulleys, couplings, applied forces, and torque inputs or outputs. Resolve forces into two perpendicular planes and calculate bearing reactions in each. From those reactions, construct bending-moment diagrams; combine the plane-specific moments at each relevant section to find the resultant bending moment. Plot torque separately, because torque may change across the shaft where power enters or leaves.

For a rotating shaft under a steady transverse load, a point on the surface cycles from tension to compression as the shaft turns through the bending field. In a stationary reference frame the force may be constant; in the material at the surface, the bending stress can be fully reversed. Torque, by contrast, may be steady in direction and magnitude, fluctuate with a process, or reverse. Treating all loads as static because their magnitudes appear constant can therefore miss the fatigue-driving cycle.

Use the real operating history where possible. A motor’s start-up torque, a jam, a reversing cycle, or a recurring impact may matter more than the nominal rated condition. Record speed, direction, load duration, starts per hour, and any known overloads. If the load spectrum is uncertain, document the assumptions and evaluate plausible severe cases rather than hiding uncertainty inside an unexplained safety factor.

For a gear force, distinguish the tangential component that transmits torque from radial and, where applicable, axial components that load bearings and the shaft. The force directions depend on gear type and arrangement. A clean load diagram prevents a common error: sizing to transmitted torque while omitting the bending caused by the gear mesh.

First-pass strength: combine bending and torsion

For a solid circular shaft of diameter d, the nominal outer-surface bending stress from a bending moment M is σ = 32M/(πd³). The nominal torsional shear stress from torque T is τ = 16T/(πd³). Use a consistent unit system: for example, newton-millimetres and millimetres produce stress in megapascals, since 1 MPa equals 1 N/mm². These equations assume a solid, circular, prismatic section and do not account for local geometric stress concentration.

When bending and torsion act together, a common preliminary check uses the distortion-energy, or von Mises, equivalent stress for a stress state dominated by normal bending stress and torsional shear:

σ_vm = √(σ² + 3τ²)

This is a useful static-yielding comparison for ductile materials under appropriate assumptions. It is not a fatigue-life calculation, and it does not make a notch disappear. If axial force is significant, include its normal stress in the stress state. If the shaft is hollow, use the section properties for the actual inner and outer diameters rather than the solid-shaft formulas.

Worked preliminary check: suppose a solid shaft section carries a bending moment of 500 N·m and a torque of 300 N·m. Convert these to 500,000 N·mm and 300,000 N·mm. At a trial diameter of 40 mm, nominal bending stress is about 79.6 MPa and nominal torsional shear stress about 23.9 MPa. The equivalent stress is approximately 89 MPa. That value is not a recommended design stress or a finished result: it excludes local concentration effects, fatigue, deflection, and uncertainty in the loads. Its purpose is to show how bending and torque combine in a first-pass calculation.

The cube dependence is significant. For a given load, nominal bending and torsional stresses vary as 1/d³. A modest diameter increase can sharply reduce nominal stress, but the real design still needs checks at the smallest or most highly stressed sections—often not at the shaft’s largest outside diameter.

Geometry turns nominal stress into local stress

At a shoulder fillet, keyseat, spline root, or groove, the local peak stress can exceed the nominal value. A theoretical stress concentration factor, K_t, describes the elastic concentration for a specified geometry and loading mode. Fatigue analysis commonly uses a fatigue stress concentration factor, K_f, which accounts for material sensitivity to the notch and is often expressed as 1 + q(K_t − 1), where q is notch sensitivity. The appropriate values depend on geometry, material, and the chosen method; they should not be guessed from a generic “keyway factor.”

Details matter because shaft features are not interchangeable. A generous shoulder radius can reduce concentration, but it must fit the bearing or gear hub. A keyway creates a notch and removes material; a spline has its own root geometry and load distribution. A retaining-ring groove may be especially unfavorable where bending is high. Locate features with the bending-moment and torque diagrams in view, and avoid placing an abrupt section change at the most heavily loaded location when the layout permits.

For a preliminary fatigue assessment, apply the appropriate fatigue concentration factors to alternating and mean stress components before combining them. The convention should be explicit: whether factors are applied to bending stress, torsional shear, or an equivalent quantity must match the method being used. A polished, uniform test specimen is not the same thing as a machined production shaft with a shoulder, surface marks, and a keyseat.

Fatigue sizing with the modified Goodman criterion

Fatigue design separates a stress cycle into alternating and mean components. For a normal stress history with maximum and minimum values, the mean stress is σ_m = (σ_max + σ_min)/2, and the alternating stress is σ_a = (σ_max − σ_min)/2. Apply the same idea to shear where needed. Rotating bending under a steady transverse load is commonly modeled as alternating bending stress with approximately zero mean at a material point, while steady torque contributes mean shear stress. Real duty cycles may be more complex.

The modified Goodman relation provides a linear preliminary boundary between alternating stress and tensile mean stress:

σ_a/S_e + σ_m/S_ut ≤ 1/n

Here, S_e is the corrected endurance limit or fatigue strength appropriate to the design method, S_ut is ultimate tensile strength, and n is the chosen design factor. The expression is a useful conceptual guide, not a license to insert raw coupon values and declare the shaft safe. The bending and torsional stress components must be handled consistently, including their mean and alternating parts and any local fatigue concentration effects. A selected multiaxial fatigue method should govern the combined-stress treatment.

The fatigue strength used in design is affected by surface finish, size, reliability target, temperature, environment, and loading mode. A machined surface, a corrosion-prone installation, and a small polished laboratory specimen do not share identical fatigue performance. Material condition and heat treatment matter as well. Where the design depends strongly on a fatigue limit, state how it was estimated and which correction factors or material data were applied.

Goodman is one of several fatigue criteria. Soderberg uses yield strength for the mean-stress intercept and is generally more conservative in that respect; Gerber uses a curved relation and may fit some ductile-metal data better. Criterion selection should follow the design context, material evidence, and required reliability. Do not treat the names as interchangeable safety labels. For variable-amplitude service, a single equivalent stress may be inadequate; a cycle-counting and cumulative-damage approach may be needed, with assumptions about the spectrum and damage rule documented.

Fatigue is often where a comfortable static result becomes a marginal design. A shaft can remain below yield on every cycle yet accumulate damage over millions of rotations. The calculation should therefore identify the most severe fatigue section, not simply report the maximum von Mises stress at the nominal operating point.

Deflection and alignment are design constraints

Strength answers whether a section is likely to yield or fatigue under modeled loads. It does not answer whether the shaft bends too far for the machine to function. Deflection can misalign gears, alter tooth contact, increase bearing edge loading, and create vibration. Shaft slope at a bearing or gear may be as important as displacement at midspan.

For a stepped shaft, calculate deformation using the actual span, support conditions, loads, and local flexural rigidity EI, where E is elastic modulus and I is the second moment of area. The familiar beam equations for a uniform span are useful for checks, but a geared shaft with several diameters and concentrated loads is better analyzed with an appropriate beam model or verified structural analysis. Model bearing supports realistically: a bearing is not automatically a perfectly fixed support, and its stiffness can influence the result.

There is no single deflection limit that suits every shaft. Gear mesh requirements, bearing type, seal alignment, coupling flexibility, and manufacturer guidance can set different limits. Compare calculated displacement and slope at functional interfaces with the limits relevant to those components. If a gear designer specifies allowable shaft slope at the mesh, that requirement is more informative than a generic statement that deflection should be “small.”

Critical speed: keep the operating range away from resonance

A shaft-and-rotor system has natural lateral vibration frequencies. When running speed approaches a lateral natural frequency, unbalance or other excitation can produce large vibration. The associated speed is commonly called a lateral critical speed. A first estimate can use a simplified rotor model, but the result depends on shaft stiffness, bearing and support stiffness, disk locations and masses, and the distribution of the shaft’s own mass. A bare-shaft calculation can miss an important gear, pulley, or coupling mass.

For a rough screening estimate of a uniform, simply supported beam with a concentrated midspan load, the static deflection δ under that load can be related to a first bending frequency by f ≈ (1/2π)√(g/δ), where g is gravitational acceleration and the units must be consistent. This approximation is useful for intuition, not a complete rotor-dynamic design method. Real machines often depart substantially from the idealized beam and point-mass assumptions.

Compare the predicted critical speed with the full operating range, including run-up and coast-down, not only the rated speed. If the machine passes through a resonance during acceleration, the transient response and control strategy may matter. If it runs near a critical speed, a more complete model and vibration review are warranted. The required separation margin depends on the system, its damping, excitation, and governing requirements; it should not be chosen by a universal rule of thumb without justification.

Critical-speed analysis also connects directly to stiffness. A diameter increase can raise bending stiffness substantially, while bearing changes or a shorter span can alter the system response. But the resulting design must still fit gears, bearings, and seals. Treating speed as a final checkbox after the drawing is released can force expensive changes.

Standards belong in the design record

Standards help establish consistent terminology, methods, and acceptance practices, but a citation alone does not verify a design. ANSI/AGMA 6001-F19 is pertinent to components for enclosed gear drives; use its published text to identify the requirements that apply to the project. Do not assume that a shaft calculation meets every requirement merely because the standard appears in a report. Record the edition, scope, and clauses or methods actually used.

For fits and dimensional tolerancing, the ISO system is another distinct design concern. Standards in the ISO 286 series define the system of limits and fits, including tolerance classes used for holes and shafts. A fit designation addresses size tolerances and the resulting clearance or interference range; it does not, by itself, specify runout, coaxiality, surface texture, or acceptable gear alignment. Those needs require their own drawing requirements and inspection methods.

Keep the strength model, material specification, manufacturing route, and inspection plan aligned. A fatigue calculation based on a particular surface condition is not meaningful if the drawing and process do not control that condition. Likewise, a fit selected from ISO limits must be compatible with the actual assembly method, service temperature, and component function. Consult current, authoritative standards rather than relying on a copied table whose edition or context is unknown.

A disciplined preliminary workflow

1. Define the duty: document power, torque, speed range, direction, starts, overloads, and expected cycles.
2. Map loads and supports: draw the shaft layout, resolve forces in two planes, and calculate reactions, bending moments, and torque by section.
3. Select trial geometry and material: account for bearing seats, gear interfaces, shoulders, keyways or splines, and realistic material condition.
4. Check nominal strength: combine bending and torsion with a suitable static criterion, then revisit the most critical local features.
5. Check fatigue: determine mean and alternating stress, apply justified notch and surface effects, and use a suitable criterion such as modified Goodman for a preliminary assessment.
6. Check function and dynamics: calculate displacement and slope at sensitive interfaces, then estimate or analyze lateral critical speed across the complete operating range.
7. Document and verify: record assumptions, standard editions, material evidence, tolerances, and inspection needs; refine the model when test data or a more detailed analysis becomes available.

This sequence is iterative. If fatigue governs at a keyseat, a larger diameter may help, but moving the feature, changing the connection, or improving the shoulder geometry may be more effective. If deflection governs, changing span or support stiffness may outperform indiscriminate material substitution. If critical speed is too close to operation, changing mass placement or support conditions may be necessary alongside diameter changes.

What a credible shaft calculation should show

A useful calculation makes its logic reproducible. Include the shaft layout, bearing locations, gear or pulley forces, torque distribution, material properties and condition, key geometric features, and units. Identify the governing section for each check; it may differ between static strength, fatigue, deflection, and critical speed. State whether loads are nominal, peak, or fatigue-cycle values, and explain the chosen design factor rather than presenting it as a substitute for load knowledge.

Finally, distinguish preliminary sizing from release-ready design. Actual dimensions depend on the load spectrum, geometry, material, manufacturing quality, service environment, and component limits. Where consequences of failure are serious or uncertainty is high, independent review, detailed analysis, and appropriate testing may be necessary. The strongest shaft design is not simply the one with the largest diameter; it is the one whose assumptions, margins, interfaces, and verification plan match the machine it must serve.