S-N curve

1 Fatigue basics and the meaning of “S” and “N”

An S–N curve, or stress–life curve, summarizes how cyclic stress levels relate to fatigue life. “S” denotes a stress measure used in testing and analysis—commonly the stress amplitude or an equivalent alternating stress component. “N” denotes the number of load cycles to a defined fatigue life endpoint, such as crack initiation or final failure.

1.1 Cyclic loading concepts

Fatigue arises when a component experiences repeated or fluctuating stresses that can initiate damage and progressively degrade load-bearing capacity. The defining feature is cyclic action: even when peak stress is below static strength, repeated cycling can lead to cracking and eventual failure.

Key distinctions include whether the stress history is steady or variable, whether cycling is fully reversed or biased toward tension or compression, and whether loading involves stress-controlled or strain-controlled conditions. These factors affect both how test data are generated and how results are interpreted in design.

1.2 Stress amplitude and stress ratio (R)

In practice, the stress measure “S” is chosen to represent the dominant cyclic portion of the loading. For many classical S–N approaches, the relevant independent variable is the alternating component (often represented by an amplitude), while mean stress is handled via a stress ratio correction.

The stress ratio “R” captures the relation between the minimum and maximum stress in a cycle. It is central because the same amplitude can produce different fatigue responses depending on whether the cycle is biased toward tensile or compressive states.

1.2.1 Alternating stress vs mean stress

A cyclic stress history can be decomposed into a mean (average over a cycle) and an alternating component (the cyclic fluctuation about that mean). Fatigue performance is influenced by both: mean stress can promote or hinder damage depending on the material and loading mode.

Common engineering practice is therefore to specify “S” for amplitude-dependent behavior while applying correction relationships to account for mean stress effects associated with different stress ratios.

1.2.2 Common definitions of life (initiation vs failure)

The “N” in an S–N curve must be tied to a defined endpoint. Two frequently used definitions are:

• Crack initiation life: the cycle count to the formation of a detectable crack, often from a surface defect or notch root.
• Failure life: the cycle count to fracture or to a predetermined loss of load-carrying capacity.

Although initiation and failure lives are correlated, they can differ significantly for the same stress level. This is especially true when crack propagation occupies a substantial portion of the overall life.

1.3 Cycle counting and fatigue life units

Fatigue data are commonly reported as the number of cycles to the chosen life endpoint, “N.” In tests, cycles may be counted based on machine-driven periodic loading. When loading is not strictly periodic, specialized cycle counting methods are used to estimate an equivalent number of stress cycles at relevant amplitudes.

The time required for a given cycle count depends on test frequency and waveform, but the fatigue life itself is expressed in cycles rather than time. Standardizing frequency and waveform during experiments is important for comparability across datasets.

2 What an S-N curve shows

An S–N curve provides a relationship between cyclic stress level and fatigue life. It condenses complex material behavior into an empirical trend derived from experiments under controlled test conditions.

Interpretation depends on which stress definition is used, the stress ratio, the life definition, and the surface and geometry conditions of the test articles. Without matching these conditions, direct application to design can lead to misprediction.

2.1 Typical curve shapes across life regimes

S–N curves are often displayed on logarithmic axes, producing approximately straight segments over portions of the life range. Many materials show different mechanisms dominating at different life scales, which leads to characteristic changes in slope.

2.1.1 High-cycle fatigue region

In the high-cycle fatigue (HCF) region, failures typically occur after a large number of cycles. Plastic strain effects at the crack initiation site are usually limited, and the behavior is often treated as predominantly elastic. The slope of the curve in this region reflects how sensitive fatigue life is to amplitude changes under small cyclic plasticity.

2.1.2 Low-cycle fatigue region

In the low-cycle fatigue (LCF) region, failures occur at comparatively smaller cycle counts. Cyclic plasticity becomes more significant, and the fatigue response is strongly influenced by strain level rather than only stress amplitude. Traditional S–N curves can still be used, but their interpretation may be less robust if plastic strain dominates the damage process.

2.2 Failure criteria used in curve construction

The endpoint used for “N” must be consistent across tests that contribute to a single curve. Failure can be defined by:

• fracture detection (complete separation),
• a load drop criterion,
• a crack length threshold associated with a standardized measurement procedure,
• or a crack initiation detectability criterion.

The use of consistent monitoring techniques affects scatter and comparability. For welded structures, additional criteria may be used to define what constitutes failure in the presence of complex crack paths.

2.3 Reading and interpolating an S-N curve

Reading an S–N curve involves mapping a design stress measure to an estimated number of cycles. Engineers typically interpolate between plotted points or fitted model lines on log–log coordinates when appropriate.

Extrapolation beyond the range of test data is a common risk. Since S–N relationships can change slope between regimes, extrapolated life predictions can be unreliable, particularly if the design stress corresponds to a life region not represented in the experiments.

3 Experimental determination of S-N curves

S–N curves are constructed from fatigue tests that systematically vary cyclic stress levels and record the corresponding life outcomes. A reliable curve depends on specimen preparation, controlled loading, accurate instrumentation, and consistent failure definitions.

3.1 Test specimens and preparation

Specimen geometry and surface condition strongly influence fatigue results. Standardized specimens may include smooth cylindrical samples, notched specimens, or specimens representing weld details. Surface finish, machining marks, and any residual defects can control crack initiation locations and thus shift the curve.

Preparation also includes material characterization and control of heat treatment. If microstructure varies across batches, it can increase scatter and reduce the statistical confidence in fitted parameters.

3.2 Loading methods and instrumentation

Fatigue tests impose cyclic loading under controlled stress or controlled strain conditions, using servo-hydraulic or mechanical test systems. Instrumentation may include load cells, displacement transducers, crack gauges, extensometers, and environmental monitoring.

3.2.1 Axial, bending, and torsional fatigue setups

Different loading modes influence stress distribution and crack orientation. Common setups include:

• Axial loading, which produces comparatively uniform stress along the gauge length.
• Bending fatigue, where stress varies linearly through a section, often creating strong surface gradients.
• Torsional fatigue, where shear stresses dominate and cracking often follows shear-related planes.

Each setup requires calibration and careful control of waveform to ensure the intended stress amplitude and stress ratio are achieved at the specimen.

3.3 Data acquisition and cycle verification

Tests typically record the number of cycles until the defined endpoint. Cycle verification ensures that the machine’s control system corresponds to the target waveform frequency and that the stress ratio remains within acceptable bounds throughout the test.

In many setups, data logging captures changes in load amplitude or mean load drift, which can occur due to stiffness changes as cracks grow. Accounting for such effects improves the consistency of the plotted results.

3.4 Building the dataset for curve fitting

A dataset for S–N curve construction includes multiple stress levels, with several specimens at each level to estimate scatter. Each specimen contributes a fatigue life “N,” and outliers may be reviewed based on documented test anomalies.

Curve fitting converts discrete experimental outcomes into a continuous relationship, often using regression on log-transformed variables and incorporating censored data when not all tests reach the endpoint (for example, runouts beyond a maximum number of cycles).

4 Curve models and fitting approaches

Because the underlying relationship is empirical, curve models must balance simplicity, interpretability, and fidelity to data. Fit choices can materially affect predicted life, especially at the boundaries of the experimental domain.

4.1 Baseline power-law (linear-in-log) models

A common model uses a power-law form, which becomes linear when plotted on logarithmic axes. In this representation, a slope parameter captures how quickly life decreases as stress increases, while intercept terms calibrate the overall life level for the dataset.

This approach can describe limited regions well, particularly where a single dominant mechanism operates and the slope does not change dramatically.

4.2 Bilinear and piecewise representations

Some materials and details exhibit changes in slope between high-cycle and low-cycle regions or between different damage mechanisms. Bilinear or piecewise models allow separate slopes in different life ranges, often improving accuracy near transition points.

Piecewise approaches require careful definition of breakpoints to avoid overfitting. They also benefit from having test coverage across the transition.

4.3 Incorporating runouts and censored data

Not all tests necessarily reach the defined failure endpoint; tests may be terminated after a maximum cycle count to save time or when a specimen remains unfailed (runout). Such data are “censored” because they provide a lower bound on life rather than the exact failure cycle count.

Statistical techniques for censored data incorporate these bounds into parameter estimation. Doing so prevents the model from being biased toward shorter lives due to selectively failing specimens.

4.4 Statistical scatter and confidence levels

Fatigue life exhibits significant scatter due to microstructural variability, surface defects, measurement uncertainty, and random differences in crack initiation sites. Accordingly, fitted curves are often accompanied by statistical descriptors such as characteristic life or confidence bounds.

Confidence levels help communicate how conservative a design prediction is relative to the underlying population of specimens. Without acknowledging scatter, point estimates can be misleading in reliability-oriented applications.

5 Mean stress and stress ratio corrections

Mean stress modifies fatigue behavior because it changes the effective driving force for crack initiation and growth during the cyclic process. Stress ratio corrections attempt to translate fatigue data measured at one stress condition to another.

5.1 R-parameter effects on fatigue life

For a given alternating stress amplitude, a tensile mean typically increases fatigue damage risk compared with a fully reversed cycle. Conversely, compressive mean can suppress certain damage mechanisms, sometimes shifting life upward.

The effect depends on material sensitivity and on whether plasticity or elastic mean effects dominate. As a result, correction accuracy can vary across metals and structural categories.

5.2 Walker and similar correction concepts

Walker-type concepts propose an empirical relationship between mean stress (or stress ratio) and the fatigue strength. Typically, a material-specific exponent governs how the correction scales with mean stress level. These formulations are often used because they remain practical and can be calibrated with limited additional tests.

In application, the correction requires using data that match the intended life definition and the same stress measure definition as the base S–N curve.

5.3 Goodman and Gerber-type approaches (conceptual use)

Goodman and Gerber-type approaches are conceptual frameworks for correcting fatigue strength based on mean stress. They describe different functional relationships between mean stress and allowable alternating stress, with parameters tied to material static properties.

Although these relations originated in older empirical contexts, they remain widely referenced because they provide straightforward methods to adjust mean stress effects for engineering estimates.

5.4 Selecting corrections for design contexts

Design practice typically selects correction methods based on:

• compatibility with the base S–N curve source,
• similarity of test conditions to the component’s loading,
• and availability of calibration parameters.

Because no single correction universally fits all materials and geometries, engineers choose the approach that best aligns with documented standards or test evidence. When uncertainty is high, more conservative strategies or reliability factors are commonly added.

6 Design application of S-N curves

S–N curves are used to estimate fatigue life for components subjected to cyclic loading. The workflow typically involves determining an appropriate design stress, selecting a relevant curve and corrections, and accounting for variability and uncertainty.

6.1 Estimating fatigue life from design stress

Given a cyclic stress amplitude at the critical location, designers map that stress onto the selected S–N curve to estimate cycles to the chosen life endpoint. For welded or notched components, the “stress at a critical point” may require special modeling or stress concentration considerations.

The stress amplitude used in the mapping must match the curve’s definition (nominal, local, alternating component, and associated stress ratio). Misalignment of definitions is a common source of error.

6.2 Converting service load spectra to equivalent cycles

Real service loading often varies in amplitude and sequence rather than remaining constant. To use an S–N curve, load spectra are commonly converted into an equivalent representation—such as an equivalent number of cycles at a reference stress amplitude using cumulative damage concepts.

This process involves establishing ranges, mean levels, and cycle counts from measured or modeled operational loads, then translating them into a fatigue-relevant cycle spectrum.

6.3 Use of safety factors and reliability targets

Predicted fatigue life is rarely treated as a single deterministic value. Instead, design uses safety factors and, in reliability-based contexts, targets a probability level for survival over a specified duration.

Safety factors reflect uncertainties including material variability, model simplifications, measurement errors in stress estimates, and discrepancies between laboratory specimens and real structures.

6.4 Handling uncertainty in material and loading

Uncertainty handling may include conservative assumptions in stress modeling, bounding choices in curve selection, and careful treatment of surface and geometry. When actual loading spectra differ from assumptions, life predictions can shift materially.

Documented assumptions and sensitivity checks help establish whether design conclusions are robust, especially when the estimated life is near the threshold of allowable service periods.

7 Materials, surfaces, and structural features

Fatigue performance is strongly governed by material microstructure, surface condition, and the presence of geometric discontinuities. These factors influence where and how cracks initiate, and thus alter the effective stress–life behavior.

7.1 Material selection and microstructural influences

Different materials have different fatigue strengths and characteristic sensitivities to microstructural features such as grain size, inclusions, and phase distributions. Even within the same nominal alloy, variations from heat treatment and processing route can produce meaningful changes in the S–N response.

Material selection also includes considering whether the dominant damage mechanisms are crack initiation from microstructural heterogeneities or initiation from surface defects.

7.2 Effect of surface roughness and finish

Surface roughness increases the likelihood of crack initiation by creating stress concentrators and local hot spots under cyclic loading. The direction and severity of surface marks from machining or manufacturing can influence both initiation life and scatter.

Surface treatments that improve finish or introduce beneficial compressive stresses can shift the S–N curve toward longer life, but the degree of improvement depends on the treatment process and depth of the altered layer.

7.3 Influence of heat treatment and residual stress

Heat treatment changes hardness, strength, and microstructural arrangement, affecting fatigue resistance. Residual stresses—introduced during manufacturing, forming, machining, welding, or heat treatment—can alter effective mean stress at the material microstructure level.

Because residual stress fields can relax during cyclic loading, their influence may be time- or cycle-dependent, adding another layer of uncertainty when translating laboratory results to service.

7.4 Welds and notches (general treatment)

Welded joints and notched geometries commonly create localized stress concentration and complex crack paths. Fatigue performance for these features is often addressed using specialized detail categories or notch-aware approaches rather than applying smooth-specimen curves directly.

General treatment includes accounting for effective notch geometry, weld toe conditions, potential undercut or surface irregularities, and the interaction between weld reinforcement shape and cyclic stress distribution.

8 Units, conventions, and common interpretation pitfalls

Correct interpretation of S–N curves requires attention to units, axis conventions, stress definitions, and test termination criteria. Small inconsistencies can lead to large errors in predicted life.

8.1 Stress measures: nominal vs local stress

Some S–N curves are based on nominal stress computed from global geometry, while others use local stresses that account for stress concentrations. Not using the same stress basis as the curve can misalign the meaning of the independent variable “S.”

For notched components, local stress can be estimated using finite element analysis or stress concentration factors, but the method must match the curve’s intended stress definition.

8.2 Diagram axes conventions (log–log vs semi-log)

Many S–N charts use logarithmic scaling for both stress and life, typically producing straight-line behavior suitable for power-law interpretation. Some presentations may use semi-log axes or different scaling, which changes the perceived slope and how interpolation should be performed.

Misreading the axes can cause errors when selecting stress values for life calculations or when comparing datasets from different sources.

8.3 Runout criteria and test termination choices

Runout is the condition where a specimen survives beyond a predefined maximum number of cycles. How runouts are treated in curve construction affects the fitted model, especially at high life values.

If a curve author uses a different runout definition or censoring strategy, direct application to design with a different termination philosophy can lead to inconsistent predictions.

8.4 Common mistakes in extrapolation beyond data range

A frequent pitfall is extrapolating the curve to stress amplitudes or life levels far outside the tested range. Since fatigue mechanisms can shift with stress and life regime, extrapolation may effectively assume an incorrect slope or transition behavior.

Another mistake is applying mean stress corrections designed for one stress measure definition to a curve based on a different stress component or local stress approach.

9 Relation to fracture mechanics and other fatigue methods

S–N curves are part of a broader fatigue analysis toolbox. Under some conditions, fracture mechanics and strain–life methods provide additional insight or improved prediction fidelity.

9.1 When S-N is complemented by crack-growth analysis

When cracks are present or when life includes significant propagation time, crack-growth analysis can complement S–N approaches. In such workflows, initiation life estimated by S–N data can be paired with growth calculations using fracture mechanics parameters.

This can improve accuracy for components where geometry effects on crack growth are prominent or where inspection intervals require knowledge of crack length evolution.

9.2 Overview of strain–life (ε-N) comparisons

Strain–life approaches relate cyclic strain amplitude to fatigue life, particularly in regions where plastic deformation is substantial. Compared with stress–life (S–N) methods, strain–life can be more suitable for low-cycle fatigue where strain control and hysteresis behavior matter.

In practice, engineers may use ε–N methods to validate or refine predictions when the component’s stress state suggests significant cyclic plasticity.

9.3 Complementary approaches for complex loading

For complex multiaxial loading, variable amplitude sequences, or non-proportional stress histories, more advanced fatigue assessments may be used alongside or instead of a simple S–N mapping. Multiaxial criteria, spectrum-based damage models, and finite element stress field extraction are common complements.

S–N remains valuable as a starting point because it provides a structured link between stress amplitude and life, especially when loading can be reasonably represented by an equivalent cyclic stress measure.

10 Engineering workflow example

An engineering fatigue evaluation using S–N curves typically starts with defining the loading scenario, selecting a relevant curve and corrections, and computing life for the critical location, then checking margins against requirements.

10.1 Defining the loading scenario and stress spectrum

The first step is to identify the component’s service loading pattern. This may involve gathering time histories from operation, converting them into stress ranges and cycle counts, and determining whether the loading is closer to constant amplitude or variable amplitude service.

Engineers also identify the critical location, often where stress concentrations, weld toes, or geometric discontinuities exist, and estimate the local cyclic stress amplitude associated with that point.

10.2 Selecting the appropriate S-N curve and corrections

Next, an S–N curve source is selected based on material class, surface condition representation, and geometry category. If mean stress effects differ from the base curve conditions, an appropriate mean stress correction framework is applied using the relevant stress ratio or equivalent mean representation.

Selection also considers the life definition used by the curve author, ensuring that the chosen “N” aligns with the design requirement (initiation vs failure).

10.3 Computing fatigue life and checking design margins

Using the stress representation and corrections, the fatigue life is estimated. For variable amplitude loading, cumulative damage concepts convert the service spectrum into an equivalent damage measure, which can then be compared to allowable life.

Finally, the predicted life is compared against required service duration and acceptable reliability or safety margins. If the estimate falls short, the evaluation may iterate by adjusting stress estimates, refining modeling of the critical zone, or exploring design changes.

10.4 Documenting assumptions and test conditions

A complete evaluation documents the assumed stress definition, the location and method for stress extraction, the correction approach used for mean stress, and how the load spectrum was reduced to an equivalent cycling representation.

It also records test-related assumptions embedded in the selected S–N curve dataset—such as life endpoint definition, test stress ratio, and treatment of runouts—so that uncertainties and limitations are transparent for review and future updates.