A component can satisfy a static stress limit and still fail after thousands of duty cycles. That gap is where a guide to Nastran fatigue analysis becomes valuable. Fatigue analysis connects realistic operating loads, finite element stress results, material test data, and a defensible damage calculation. It is not a single solver button or a contour plot. It is a workflow that must be validated at every stage.
For engineering teams developing structures in aerospace, transportation, heavy equipment, energy, or medical devices, the objective is straightforward: identify where cyclic loading will initiate damage, estimate usable life, and make design decisions before expensive physical testing or field failures occur. Achieving that objective requires more than a refined mesh. It requires the right load representation, stress recovery method, material data, and interpretation of uncertainty.
What Nastran Fatigue Analysis Actually Requires
Nastran is exceptionally capable at generating the structural response data used by fatigue methods. Depending on the event, that response may come from linear static, modal, frequency response, transient, nonlinear, or random vibration analysis. A fatigue calculation then uses stress or strain histories, together with material fatigue properties and a damage model, to estimate life or damage.
The most common workflow is stress-life, often called S-N analysis. It is generally appropriate for high-cycle fatigue, where stresses are predominantly elastic and the expected life is relatively long. Strain-life, or E-N analysis, is more appropriate when local yielding, low-cycle fatigue, or severe notch effects govern the response. Fracture-mechanics methods address crack growth when an initial flaw is known or assumed. These approaches answer different questions and should not be treated as interchangeable.
A Nastran-based workflow may use native capabilities, a fatigue module, or a specialized post-processing application. The interface changes, but the engineering requirements do not. The analyst still needs to establish how a load is applied, what stress is extracted, how cycles are counted, and whether the model represents the physical component closely enough to support a life prediction.
Start With the Duty Cycle, Not the FE Model
Fatigue results are only as credible as the load history behind them. A static load case labeled “maximum operating load” does not describe a fatigue event by itself. The key questions are how often the load occurs, whether it reverses direction, what mean load is present, how loads are sequenced, and which operating conditions are representative of the product’s intended life.
For a bracket on off-road equipment, for example, the duty cycle may include engine vibration, repeated vertical acceleration, occasional impact events, and preload from fasteners. Treating every event as a fully reversed sine wave may be conservative in some locations and dangerously nonrepresentative in others. A reliable analysis separates these load sources and documents the assumed number of cycles for each.
Variable-amplitude histories require additional care. Time-domain loading can be reduced through cycle counting methods such as rainflow counting, which converts an irregular history into stress ranges and mean stresses. Frequency-domain methods can estimate fatigue damage efficiently for stationary random processes, but their assumptions must match the physics of the event. Random vibration is not simply a long static load case with a larger factor of safety.
When measured field data is available, use it to anchor the loading spectrum. When it is not, establish a traceable spectrum from mission requirements, test specifications, operational envelopes, and engineering judgment. The uncertainty should be visible in the report rather than hidden in an arbitrary multiplier.
Build a Model That Produces Usable Stress Results
The FE model must be suitable for both global load transfer and local stress recovery. These objectives can compete. A coarse global model may predict overall stiffness and reaction loads well, yet miss the local stress gradient near a fillet, weld toe, fastener hole, or contact transition where fatigue damage begins.
Begin by checking the fundamentals: connectivity, shell offsets, material orientation, joint representation, constraints, mass distribution, and load paths. Then review whether the model includes the features that control the fatigue hot spot. A small radius omitted from a solid model, or a connection idealized as perfectly rigid, can change the local result enough to make a calculated life meaningless.
Mesh refinement should be driven by stress convergence at the fatigue-critical location, not by a target element count. Compare stress results across practical mesh densities and confirm that the reported quantity converges. This is particularly important near geometric discontinuities. The highest element stress at a sharp re-entrant corner may be a mathematical singularity, not a physical fatigue prediction.
For shell structures, the analyst must also choose the correct surface stress and account for shell orientation. For solid models, evaluate stress at locations that correspond to the physical surface where cracking is expected. Element-center values, nodal averages, and extrapolated stresses can produce materially different fatigue inputs. Consistency between the stress extraction method and the material or fatigue-method calibration is essential.
Select the Right Nastran Response Analysis
The fatigue event determines the Nastran solution sequence and response data required. A linear static analysis, commonly SOL 101, can support proportional loading when the cyclic stress state scales predictably with load. Modal and frequency response analyses are often used for harmonic excitation or vibration-driven fatigue. Direct transient or modal transient analyses are appropriate when time sequencing, shocks, and changing loads affect the response.
Nonlinear behavior requires special caution. Contact opening and closing, plasticity, large deformation, and nonlinear fastener behavior can make a simple linear stress scaling invalid. A nonlinear transient analysis may be necessary to obtain realistic stresses, but the fatigue method must also be capable of handling the resulting local response. In many cases, this is where an elastic stress-life workflow should give way to strain-life analysis, local submodeling, or focused physical testing.
Modal participation deserves a specific check in dynamic fatigue work. An analysis that omits a mode contributing to local stress can substantially overpredict life even if global displacement plots look reasonable. Review the frequency range, modal effective mass, damping assumptions, and stress response around resonances. Damping is often one of the least certain inputs and one of the most influential.
Apply Material Data and Mean-Stress Corrections Carefully
Fatigue material data is not generic strength data. An ultimate tensile strength and a yield strength do not replace an S-N or E-N curve. The selected curve should reflect the material specification, heat treatment, surface condition, product form, temperature, and environment as closely as practical.
Mean stress matters because a tensile mean stress generally reduces fatigue life while a compressive mean stress can improve it within limits. Goodman, Gerber, and other mean-stress correction approaches make different assumptions. There is no universal best choice. The appropriate correction depends on the available test data, material behavior, internal standards, and the level of conservatism required for the application.
Welded structures require their own treatment. Fatigue performance at a weld is commonly governed by weld geometry, residual stress, and fabrication quality rather than the parent material’s polished specimen curve. Use a method and fatigue class intended for welds, and define whether the analysis uses nominal stress, hot-spot stress, or an effective notch approach. Applying base-material S-N data directly to peak stress at a weld toe is a frequent and costly mistake.
Calculate Damage, Then Challenge the Result
For multiple load blocks, cumulative damage is often estimated using Miner’s rule: each cycle block contributes a fraction of damage, and the fractions are summed. A damage value near 1.0 is generally interpreted as expected initiation at the target life. The method is practical and widely used, but it does not fully capture sequence effects, overload interactions, or every material phenomenon.
That limitation does not make the calculation useless. It means the result should be presented as an engineering prediction bounded by assumptions. Report the critical locations, governing events, stress ranges, mean stresses, material curve source, correction factors, mesh sensitivity, and target mission definition. A life contour without this context is difficult to review and easy to misuse.
Correlation closes the loop. Compare predicted strain, natural frequencies, static deflection, and failure locations against test data whenever possible. Test correlation does not require a perfect numerical match; it requires an explained relationship between model behavior and physical behavior. If a prototype cracks in a different location than the model predicts, investigate the load path, weld detail, residual stress, manufacturing variation, and boundary conditions before adjusting the fatigue curve.
Common Failure Modes in the Workflow
Several errors recur across fatigue programs: using singular peak stresses as direct inputs, applying polished-specimen material data to rough or welded production parts, neglecting preload and mean stress, assuming a nominal load represents a duty cycle, and accepting dynamic results without checking modal coverage. Another common issue is treating a fatigue life result as more precise than its inputs justify.
A useful review asks whether the predicted critical location is physically plausible, whether the stress quantity matches the chosen fatigue method, and whether a reasonable change in load or material assumptions changes the decision. Sensitivity studies are often more valuable than adding another decimal place to a life estimate.
Experienced Nastran support can shorten this validation cycle, especially when a program combines advanced connections, nonlinear behavior, dynamic loading, or specialized fatigue post-processing. eNastran Engineering works with teams that need both solver-level understanding and practical judgment about what the model can legitimately claim.
The best fatigue analysis is not the one with the most colorful contour plot. It is the one that gives the design team a traceable reason to change a radius, revise a weld detail, reduce a load, alter a material process, or proceed confidently to test.