Designing Rotating Members: Torsion, Bending, and Critical Speed

The first calculation is only the beginning

For a solid circular shaft under pure torsion, the maximum nominal shear stress is τ = 16T/(πd³). Here, torque T is measured in newton-metres (N·m), diameter d in metres (m), and stress τ in pascals (Pa). The cubic dependence on diameter is the striking part: a modest increase in diameter can sharply reduce stress, while a small reduction can have a disproportionate effect.

Take a solid steel member with a diameter of 20 mm carrying 50 N·m. Converting the diameter to 0.020 m gives a nominal torsional shear stress of approximately 31.8 MPa. That is a useful first result, not a complete verdict. Real shafts may also bend under gear forces, carry rotating or reversing loads, and contain shoulders, keyways, grooves, or cross-holes that amplify local stress.

That distinction matters in practice. A shaft does not experience its drawing: it experiences the force path through the assembled machine. The motor, coupling, pulley, gear, bearings, and driven load determine which forces reach each section. Design begins by tracing those forces, then checks strength, stiffness, fatigue, and dynamic behaviour rather than treating one stress number as a pass-or-fail answer.

Torque, diameter, and torsional stiffness

Torque is the turning effect of a force. When a force acts perpendicular to a lever arm, its magnitude is T = F × r, where force F is in newtons (N) and distance r is in metres (m). A 100 N tangential force acting 0.5 m from an axis, for example, produces 50 N·m. In a shaft, torque creates shear stress and twists one cross-section relative to another.

For a uniform solid circular shaft, the polar second moment of area is J = πd⁴/32, in m⁴. The maximum elastic shear stress can also be written τ_max = Tc/J, with c = d/2. Substituting the circular-section expression for J produces the familiar formula above. For a hollow shaft with outside diameter D and inside diameter d_i, J = π(D⁴ − d_i⁴)/32. Removing material near the axis can preserve much of the torsional stiffness, because material farther from the centre contributes more strongly to J.

Stress is not the only torsional criterion. The angle of twist of a uniform shaft is θ = TL/(JG), where length L is in metres and shear modulus G in pascals; θ is in radians. The expression makes the design trade-off plain: a longer shaft twists more, while a larger polar moment or stiffer material reduces twist. In a precision drive, excessive angular deflection can disturb timing or positioning even if the material remains comfortably below yield.

For the 20 mm example, steel with a representative shear modulus of 80 GPa, a 0.5 m length, and 50 N·m torque twists by about 0.124 rad, or 7.1 degrees, if the whole length is uniform and the torque is constant. That is not a final assembly prediction: couplings, bearings, stepped diameters, and boundary conditions change the result. But it reveals why a stress-only check can miss a meaningful stiffness problem.

Bending enters through the machine around the shaft

Torque often arrives with bending. A belt pulls on a pulley; a gear tooth pushes on its mate; a rotor’s weight loads the bearings; a misaligned coupling can impose additional forces. These loads create bending moments along the shaft. In a solid circular section, the nominal maximum bending stress is σ_b = 32M/(πd³), where bending moment M is in N·m and stress is in Pa.

Consider a simply supported shaft with a central transverse load W. For an ideal point load, the largest bending moment is M_max = WL/4, where support span L is in metres. The real load may not sit at midspan, and a shaft with multiple gears or overhung pulleys needs a piecewise beam calculation. Still, this simple case demonstrates why span is consequential: increasing the distance between supports raises the bending moment for a given central load.

For a circular shaft, bending normal stress and torsional shear stress act together. Under a static, ductile-material assessment, a common comparison is von Mises equivalent stress:

σ_v = √(σ_b² + 3τ²)

This equation is a useful screening tool for combined normal and shear stress in a simple state of loading. It does not automatically account for fatigue, local stress concentration, residual stress, brittle behaviour, or every multiaxial loading condition. Nor does it replace a clear free-body diagram: the bending moment must be found at the section being checked, and the torque may vary from one side of a gear or coupling to the other.

A practical analysis often starts with two perpendicular bending planes. Resolve forces in each plane, draw shear-force and bending-moment diagrams, and combine the bending moments at each section using M = √(M_y² + M_z²). Then combine bending and torsion. This is especially important when a gear’s radial and tangential forces act in different directions and the shaft rotates under a stationary transverse load.

A worked screening example

Suppose the 20 mm solid steel shaft carrying 50 N·m also experiences a bending moment of 40 N·m at the section of interest. Assume, for this first pass, a smooth circular section with no keyway and elastic, static loading. Its nominal bending stress is:

σ_b = 32(40)/(π × 0.020³) ≈ 50.9 MPa.

The torsional stress remains approximately 31.8 MPa. Combining the two with the von Mises expression gives:

σ_v = √(50.9² + 3 × 31.8²) ≈ 75.2 MPa.

The result is not a recommendation to use a particular grade of steel. A material’s yield strength depends on its specification, heat treatment, product form, and condition. The design also needs an appropriate safety factor and allowance for the actual geometry, loading uncertainty, manufacturing, and environment. The calculation instead shows how bending changes the assessment: neither the torsion-only value of 31.8 MPa nor the bending-only value of 50.9 MPa describes the combined nominal state.

Diameter remains a powerful lever. At a fixed torque and bending moment, both nominal stresses in a solid round shaft vary as 1/d³. Increasing the diameter from 20 mm to 25 mm would multiply those stress estimates by (20/25)³, or 0.512. In other words, the idealized stresses would fall by roughly 49 percent. But the larger diameter may require different bearings, seals, couplings, or gear hubs, and it can add mass and alter the system’s vibration characteristics. A calculation must stay connected to the actual assembly.

The notch can govern the result

Nominal formulas describe an ideal, smooth section. Machines are built with transitions and features: a shoulder locates a bearing, a keyway carries a hub, a snap-ring groove retains a component, and a cross-hole may secure a fastener. Such features disturb the local stress field. The local peak can be higher than the nominal stress, sometimes substantially, depending on the feature’s shape, dimensions, orientation, and loading.

A generous shoulder fillet is therefore more than a cosmetic refinement. A sharp corner concentrates stress; a larger, smoothly blended radius generally reduces that concentration. The radius must still fit the mating bearing or component, and the drawing should specify a geometry that can be manufactured and inspected. The fillet also cannot be treated in isolation: bearing chamfers, assembly clearances, and the location of the load-bearing shoulder all matter.

For preliminary work, an elastic stress-concentration factor K_t may be applied to a nominal stress to estimate a local elastic peak, using geometry-specific charts or validated analysis. Fatigue calculations use a fatigue notch factor K_f, which may differ because materials do not respond to notches exactly as an ideal elastic model predicts. A keyway can be particularly consequential when bending stress reverses at the material point as the shaft rotates.

Not every feature should be assigned a guessed multiplier. Use a recognized design method, reliable geometry data, or finite-element analysis with an appropriate mesh and boundary conditions. A detailed simulation with an unrealistic sharp corner or poorly represented contact can produce impressive but misleading precision. For critical machinery, the analysis should be checked against standards and reviewed by a qualified engineer.

Fatigue: the load may be modest, but it keeps returning

A shaft that survives one application of a load can fail after millions of cycles. Fatigue concerns repeated stress variation, and rotating shafts create a characteristic trap: a bending load fixed in space can produce alternating bending stress at a point on the rotating surface. As the shaft turns, that material point moves from the tensile side of the bend to the compressive side, even if the external force itself never changes direction.

Torque may be steady, pulsating, reversing, or irregular. A motor start, abrupt stop, impact, or reciprocating mechanism can make the torsional component fluctuate. Engineers distinguish mean stress from alternating stress, account for material fatigue data and surface condition, and consider size, reliability, temperature, and notch effects. The relevant fatigue approach depends on the duty cycle and the consequences of failure; a single static safety factor does not represent the whole problem.

Surface finish and manufacturing details influence fatigue performance. Tool marks, corrosion pits, poor transitions, and damage during assembly can become initiation sites. Heat treatment may improve some properties while also introducing residual stresses or distortion that require control. A keyway’s geometry and finish matter, as do the fit and alignment of components that might impose unexpected cyclic loads.

Fatigue assessment should use a defined loading history where possible. Record operating speed, start-stop frequency, torque peaks, direction changes, and any known imbalance or shock. If loads are uncertain, make that uncertainty explicit rather than hiding it inside an unjustified material choice. A more conservative assumption, a monitored prototype, or a refined measurement may be warranted depending on risk.

Critical speed is a vibration question, not a strength limit

A shaft can be below its stress limit and still vibrate dangerously. A rotor-bearing system has natural frequencies. When running speed approaches a lateral natural frequency, an imbalance or other excitation can produce a large response. The corresponding resonant operating speed is often called a critical speed. Real systems have several modes, and the first lateral critical speed is only one part of the dynamic picture.

An idealized estimate for a simply supported, uniform beam with a central point mass can be built from its static deflection. If the central mass m dominates the shaft’s distributed mass, the approximate lateral stiffness at midspan is k = 48EI/L³, where Young’s modulus E is in Pa, second moment of area I in m⁴, and span L in m. The corresponding natural angular frequency is ω_n ≈ √(k/m) in rad/s, and frequency is f_n = ω_n/(2π) in hertz (Hz). The associated speed in revolutions per minute is n = 60f_n.

These are model equations, not universal shaft formulas. Bearing flexibility, shaft mass, rotor position, overhang, gyroscopic effects, damping, and support stiffness can all shift the frequency. A more complete rotor-dynamic model represents the distributed shaft and attached components, along with bearing properties. The simple estimate is most useful for seeing how design variables influence the trend.

For a solid circular section, I = πd⁴/64. Since the beam stiffness above is proportional to EI/L³, increasing diameter can strongly increase lateral stiffness; increasing span reduces it sharply. If a central mass dominates and other assumptions stay fixed, the natural frequency varies approximately as d²/L^3/2, because stiffness scales with d⁴/L³ and frequency scales with the square root of stiffness. A small change in support span can therefore matter as much as a material upgrade—or more.

To illustrate the scale, consider a 0.5 m span of 20 mm diameter steel with a 2 kg point mass at its centre, using E = 200 GPa and ignoring shaft mass and bearing flexibility. The beam model gives a stiffness of about 30,200 N/m and a natural frequency near 19.6 Hz, equivalent to approximately 1,180 rpm. This is only a teaching estimate; a real rotor would need a more representative model. It does show why operating-speed checks cannot be inferred from stress calculations.

Engineers normally consider the full operating range, transient passage through resonance, imbalance, damping, and separation between operating speeds and critical speeds. There is no single safe margin that applies to every machine: the appropriate criteria depend on the rotor, its controls, bearing type, duty, and governing design requirements. For high-speed or safety-critical equipment, rotor-dynamic analysis and applicable industry standards are essential.

Material choice is a system decision

Choosing a stronger alloy can raise the allowable stress, but strength is only one property. Elastic modulus controls stiffness in the linear range, and common steels have broadly similar Young’s moduli despite differing considerably in yield strength. A stronger steel may therefore improve resistance to yielding without delivering the stiffness increase a designer hoped for. For a beam-like shaft, geometry often has more leverage over bending stiffness than a modest change in steel grade.

Other properties still shape the decision: fatigue resistance, toughness, corrosion resistance, heat-treatment response, machinability, availability, and cost. A shaft working in a wet environment may need corrosion protection or a different material strategy; one with a press-fit hub may need attention to local stresses and assembly conditions. Material data must match the actual condition and product form, not merely a generic alloy name.

Mass is another consideration. A larger or denser rotor can increase bearing loads and alter dynamic behaviour. Hollow sections can reduce mass while retaining useful bending and torsional stiffness, but the wall must withstand local stresses, manufacturing limits, connections, and potential damage. The best section is not automatically the lightest, strongest, or thickest one: it is the section that meets the system’s constraints with a defensible margin.

A practical sequence for design review

1. Map the machine. Identify supports, gears, pulleys, rotors, couplings, overhangs, and load directions. Separate forces into the relevant planes.
2. Find torque and reactions. Use power and speed, when appropriate, with P = Tω, where power P is in watts (W) and angular speed ω in rad/s. Include transient and reversing loads where they occur.
3. Build moment diagrams. Determine bending moments and torque at each critical section rather than assuming one value applies along the full shaft.
4. Check nominal strength and twist. Evaluate bending, torsion, combined stress, and angular deflection. Use consistent SI units throughout.
5. Inspect geometry. Review shoulders, fillets, keyways, retaining grooves, threads, and fits. Apply suitable concentration and fatigue methods rather than relying only on smooth-bar equations.
6. Check service life and dynamics. Estimate fatigue under the duty cycle, calculate deflection, and assess critical speeds and vibration response across the operating range.
7. Validate assumptions. Compare the model with the drawing, measured loads, manufacturing process, and applicable standards. For consequential applications, obtain qualified engineering review and testing.

The useful answer is a balanced one

Designing a rotating member is an exercise in balancing failure modes. Torsional shear stress answers how torque loads the section; bending stress describes transverse loading; combined-stress checks assess their interaction; fatigue asks what repeated cycles will do; and critical-speed analysis asks how the assembled rotor behaves dynamically. Diameter can improve several of these measures at once, but it also changes interfaces, mass, cost, and vibration characteristics.

The 20 mm steel example makes the central point concrete. At 50 N·m, its nominal torsional shear stress is about 31.8 MPa. Add a 40 N·m bending moment and the simple von Mises estimate rises to about 75.2 MPa—before accounting for a keyway or shoulder. Meanwhile, span and support conditions may place a natural frequency near a speed the motor actually uses. Those are separate checks, and none can be safely substituted for another.

Good design is not simply choosing a tougher material or adding metal until the numbers look comfortable. It is understanding where loads travel, which details concentrate stress, how the part flexes, and how it responds over time. When those questions are answered together, a shaft becomes more than a cylinder with a torque rating: it becomes a carefully considered part of a working mechanical system.