💥 1. Why Machines Die of Fatigue, Not of Strength
New designs are checked against yield strength, and most stress analyses end at the static check: the highest load divided by the cross-section stays below the allowable stress, so the part must be safe. Then the machine goes into service, the load oscillates ten million times, and a clean-looking shaft snaps at a fraction of the yield strength. That is fatigue, the silent failure mode that takes more machines out of service than any overload ever does.
Fatigue is a micro-scale story. Repeated loading nucleates tiny cracks at stress concentrations, on the surface, or at inclusions; each cycle grows the crack a little; and the crack finally reaches a critical size where the remaining section fails fast. The engineering lesson is that fatigue life is governed by the alternating stress range, not the peak stress, and by the geometry that concentrates it, not just the material strength. This article builds the fatigue toolkit: the S-N curve, the stress-life and strain-life approaches, the damage accumulation rules, and the crack growth logic that turns a fatigue guess into a life prediction.
📉 2. The S-N Curve: The Database of Survival
The S-N curve is the empirical heart of fatigue design. Laboratory specimens are tested at various alternating stress amplitudes, and the number of cycles to failure is recorded on a log scale, producing a curve that falls with the cycle count. For steel, the curve famously flattens at about ten million cycles, the fatigue limit or endurance limit, below which the material is assumed to survive indefinitely. Aluminium has no such knee; its curve keeps falling, so aluminium designers must design for finite life at a stated cycle count.
Interpreting the S-N curve requires remembering what it represents: the median survival of polished, unnotched, axially loaded specimens. Real parts carry notches, surfaces, sizes, temperatures, and manufacturing processes that all shift the curve down. The applied fatigue correction factors, for geometry, surface finish, size, loading mode, temperature, and reliability, are the practical engineering of fatigue, and skipping them is how textbook steels fail in service.
🛠️ 3. Stress-Life vs Strain-Life: Two Families of Predictions
The stress-life method, built on the S-N curve, counts cycles against the nominal stress amplitude and is the right tool for high-cycle fatigue, regimes above about ten-thousand cycles, where the stress stays elastic and the crack grows slowly. It is simple, well-documented, and the everyday method for shafts, gears, and structural members. Its weakness is that it ignores the local plastic flow at notches, so it struggles to predict the lives of severely notched parts loaded near yield.
The strain-life method, associated with the Coffin-Manson relationship, works at the notch root where local strain cycles in the plastic regime during low-cycle fatigue, below about ten-thousand cycles. It models total strain amplitude as the sum of elastic and plastic terms and predicts the cycles to crack initiation. It is the method of choice for pressure vessels, turbine blades, and components whose service includes overloads, and it demands a level of material data and computation that the stress-life method does not.
The choice between them is a service-regime decision, not a preference. Low-cycle, plastic-dominant loading calls for strain-life; high-cycle, elastic loading calls for stress-life; and any design that crosses both regimes should acknowledge the transition and the extra uncertainty it carries.
🧩 4. Miner’s Rule and Damage Accumulation: Real Loads Are Not Constant
Real machines rarely see a single constant amplitude. A car axle suffers a spectrum of road loads; a crane hoist cycles through empty and fully loaded lifts; a pump shaft rides through start-up transients and steady running. Miner’s rule assembles this reality into a single damage index: each stress amplitude at its cycle count contributes a damage fraction equal to the count divided by the life at that amplitude, and the part fails when the sum of these fractions reaches one.
Miner’s rule is simple and, for that simplicity, Box 1 of the analyst’s report card: the order of the loading sequence, which matters in reality, does not appear in the formula. It also assumes damage is linear, which real materials approximate better in some regimes than others. The practical response is to treat Miner’s result as a screening tool, apply a healthy factor, and build the cumulative damage spectrum from measured or duty-cycle data rather than from a cheerful guess at the load history.
Rainflow counting is the companion method that turns a messy load-time history into the closed stress cycles that Miner’s rule can consume. It identifies the peaks and valleys, pairs them into full cycles, and discards the rest, producing a cycle-count histogram that connects the driving of the machine to the fatigue arithmetic of the material.
🔬 5. Stress Concentrations: Where Cracks Choose to Live
The fatigue failure almost always begins at a stress concentration: a fillet too tight, a keyway edge, a hole edge, a thread root, a weld toe. The stress concentration factor multiplies the nominal stress locally, and because fatigue life is extraordinarily sensitive to stress amplitude, a factor of two at the notch can cut life by an order of magnitude. The S-N curve for the plain specimen and the S-N curve for the shouldered part can differ by factors that dwarf every other correction combined.
The notch sensitivity of the material tempers the geometry factor. A tough steel may not develop the full theoretical concentration because the local plastic zone redistributes load; a hardened or high-strength material is more notch-sensitive and pays the full price of the geometry. The fatigue notch factor combines the theoretical concentration with this material sensitivity, and it is this effective factor that enters the life calculation.
Designers can spend the concentration budget where it matters. Generous radii at every shoulder, streamlined keyways, rolled thread roots, ground surfaces at the critical section, and shot peening to put the surface in residual compression all reduce the effective stress amplitude at the site where failure would begin. These are cheap details with expensive consequences, and the fatigue analyst learns to look for them before reaching for stronger steel.
🌧️ 6. Mean Stress: The Steady Load That Raises the Tide
Most service loading is not fully reversed; it rides on a steady mean stress. A bolted joint sees a mean tension with a fluctuating component, and the mean stress shifts the whole stress cycle upward. The endurance diagrams, the Goodman, Gerber, and Soderberg lines, map how the allowable alternating stress falls as the mean stress rises. Goodman, the most conservative and most used for brittle and high-strength materials, draws a straight line from the endurance limit at zero mean stress to the ultimate strength at zero alternating stress.
The diligent analyst always asks what the mean stress is, because ignoring it transforms a safe alternating range into an unsafe one. Residual stresses compound the effect: beneficial compressive residual stress from shot peening or rolling lowers the effective mean stress and can extend endurance dramatically, while tensile residual stress from welding or grinding does the reverse. The fatigue design that ignores mean and residual stress is designing blind to the very forces it is fighting.
📊 7. Fatigue Crack Growth: When a Crack Is Already There
Fracture mechanics takes over where initiation ends. The part already carries a crack, a manufacturing flaw, a weld defect, or a service-initiated crack, and the question is whether it grows and how fast. The Paris law relates the crack growth rate per cycle to the stress intensity factor range at the crack tip: below a threshold, the crack does not grow; above the threshold, growth accelerates until the critical stress intensity where fast fracture begins.
The practical engineering is the inspection interval. Given a detectable crack size from the inspection method and a critical size at which failure is certain, the Paris growth rate dictates how many cycles the part survives between them, and the inspection schedule is set to a fraction of that life. This is the logic behind damage-tolerant design in aircraft and pressure equipment: recognize the flaw, plan the inspection, and retire the part before the crack reaches the critical dimension.
🚀 8. Life Prediction in the CAE Workbench
Finite element analysis feeds the fatigue calculation with detail the hand calculation cannot reach. The FEA stress field resolves the notch, the assembly preload, and the thermal loads, and the post-processor extracts the stress history that the S-N or strain-life method consumes. The results inherit every assumption of the analysis: the mesh at the notch, the material curve, the load spectrum, and the correction factors. The FEA fatigue life is a number with a pedigree, and the pedigree must be examined before the number is quoted.
The honest workflow couples the simulation to a validation test. Measure a strain gauge at the critical location during a duty-cycle test, compare the measured and predicted life, and close the loop on both the load model and the material model. Every fatigue prediction is a hypothesis about the process and the material, and the hypothesis earns its keep when the test agrees with it.
✅ 9. Fatigue Design Checklist
Identify the alternating stress range at the critical section, not just the peak stress, and count the real duty cycle. Use stress-life for high-cycle elastic service and strain-life for low-cycle plastic service, and state the regime. Apply the correction factors for surface, size, load, temperature, and reliability to the S-N curve. Reduce the stress concentration with generous radii, rolled roots, and ground or peened surfaces at the critical site. Account for the mean stress with the Goodman family of lines, and decide whether residual stress is friend or enemy. Build the load spectrum with rainflow counting, apply Miner’s rule with a margin, and validate the prediction with a strain-gauge and duty-cycle test. If cracks are possible, define the detectable and critical sizes and set the inspection interval from the Paris growth law. Finally, record the analysis with its assumptions and versions, because a fatigue prediction without its pedigree is a number waiting to be misquoted.
🔚 10. Conclusion
Fatigue is the accountant’s failure: it counts every cycle, and it remembers every overload and every sharp corner. The machine fails not when the yield line is crossed but when the accumulated damage of a lifetime of small cycles reaches its limit. Design for the range, respect the notch, feed the spectrum, and validate the prediction, and the part survives its service life not by luck but by the same arithmetic that predicted it.
⚙️ 11. A Worked Example: The Rolling Mill Spindle
Consider a mill spindle carrying a reversing torque that cycles a thousand times per day. The shaft shoulder, located at a fillet, sees a fully reversed alternating stress of roughly 180 megapascals on a steel whose endurance limit, after surface, size, and notch corrections, lands near 160 megapascals. The plain static check says the shaft is safe; the fatigue check says the alternating range exceeds the corrected endurance limit, and the shaft will not reach its design life.
The fix is geometric, not metallurgical. Enlarging the fillet radius from two to eight millimetres cuts the theoretical stress concentration from above two to about 1.3, raising the corrected endurance limit past the applied range, and the revised calculation predicts an effectively unlimited life in this regime. A strain gauge fitted at the fillet during a commissioning run confirms the predicted local stress within ten percent, and the shaft is released with a recorded margin. The anecdote is the whole discipline in miniature: the static check passed, the fatigue check failed, the geometry changed, the gauge confirmed, and the machine earned its service life in the drawing office rather than on the scrap heap.
📚 12. Where to Dig Deeper
The fatigue literature is deep and honest about its uncertainties. The classic design codes, such as the ASME Boiler and Pressure Vessel Code and the European fatigue design recommendations, codify the correction factors and the safety margins for structural steel. Material suppliers publish fatigue curves for their own grades, and the measured data always beat the generic textbook curves for the same steel. For the practicing designer, the discipline is to pair the code, the material data, and a validation test, and to treat the three together as one fatigue prediction system rather than three disconnected sources of confidence.
