Thermal Fatigue Analysis Using FEA: Predicting Failure Under Repeated Temperature Cycles

A component can fail without ever carrying a heavy load. Heat it, cool it, heat it again, and if its expansion is restrained - by a bolted joint, by a neighbouring material, or simply by its own cooler interior - every cycle forces the material to strain a little, then unstrain, then strain again. No external force is pushing on it. The temperature swing alone is doing the work, and eventually a crack appears. Thermal fatigue analysis using FEA is how engineers predict where that crack will start and how many cycles it will take, before the part goes into service.

Key takeaways

• Thermal fatigue is driven by restrained expansion and contraction, not external loading - the stress comes from the temperature change itself.
• The main drivers are constraint, temperature gradients through the part, and CTE mismatch between joined materials.
• It is usually a low-cycle fatigue problem involving cyclic plastic strain, so strain-life methods such as Coffin-Manson are the usual approach.
• A credible study couples a transient thermal analysis with a structural one, using temperature-dependent material properties.

Table of Contents

• How Temperature Creates Stress
• The Three Main Drivers
• Why It's Low-Cycle Fatigue
• Where It Matters Most
• The Analysis Process
• Common Mistakes
• FAQ

How Temperature Creates Stress

A free, unconstrained body expands when heated and contracts when cooled without developing any stress at all. Stress appears only when that movement is prevented. In the fully constrained case, the resulting stress is roughly the material's stiffness multiplied by its expansion coefficient and the temperature change - which is why even a modest temperature swing can produce very large stresses in a stiff, rigidly held part.

Thermal strain = CTE × ΔT  |  Restrained strain → Stress → Repeated every cycle

Constraint is what turns heat into damage: The same temperature change that harms a rigidly held component may do nothing to an identical part free to expand. FEA is useful here because it models the real constraints and geometry, not an idealised free or fully fixed case.

The Three Main Drivers

Driver	What happens	Typical example
External constraint	Expansion is blocked by supports, fasteners or adjoining structure	A pipe run held between fixed anchors
Temperature gradient	Hotter regions try to expand more than cooler ones, straining each other	A brake disc or turbine casing during rapid heating
CTE mismatch	Bonded materials expand by different amounts for the same temperature change	Solder joints, coatings, bonded multi-material assemblies

Real components usually experience more than one of these at once, which is a large part of why hand calculations struggle, and simulation earns its place.

Why It's Low-Cycle Fatigue

Thermal stresses from large temperature swings are often high enough to cause localised plastic strain on every cycle. Once plastic strain is involved, the component typically fails after relatively few cycles, and the stress-based S-N approach used for high-cycle fatigue stops being the natural tool. Instead, life is usually estimated from the strain range per cycle, using strain-life relationships such as Coffin-Manson.

Look at the hysteresis loop, not just peak stressEach temperature cycle traces a stress-strain loop at the critical location. The width of that loop, the plastic strain range, is what drives damage. Peak stress alone can hide whether the material is cycling elastically or accumulating plastic strain every cycle.

For the high-cycle side of the same discipline, see our article on FEA for Gears and Drive Components, and for how nonlinear behaviour is handled in the solver, see Large Deformation Analysis for Nonlinear Mechanical Components.

Where It Matters Most

• Engine and exhaust components - manifolds, cylinder heads and turbocharger housings cycling between cold start and full operating temperature
• Electronics packaging - solder joints and die attach layers cycling with power and ambient temperature
• Pressure vessels and piping - start-up and shutdown cycles across thick sections with steep thermal gradients
• Brake discs and friction components - rapid, repeated surface heating with a much cooler core
• Power generation equipment - turbine and boiler parts under repeated load-following cycles

The Analysis Process

Define thermal cycle → Transient thermal → Map to structural → Extract strain range → Estimate life → Validate

1. Define the thermal cycle: temperature range, heating and cooling rates, dwell times and expected number of cycles.
2. Run a transient thermal analysis: solve for the temperature field over time, capturing the gradients that develop, rather than assuming a uniform temperature.
3. Map temperatures to the structural model: apply the temperature field as a load, with temperature-dependent material properties and the real constraints.
4. Include the right material behaviour: represent plasticity, and creep where relevant, since thermal fatigue typically involves cyclic plastic strain.
5. Extract the strain range per cycle: review the response over stabilised cycles at the critical location.
6. Estimate life and validate: apply a strain-life relationship and check against test or field data where available.

The first cycle isn't the representative one: Materials often shake down or ratchet over the first few cycles. Reading the strain range from cycle one can misjudge the steady-state behaviour that actually governs life, so run enough cycles for the response to stabilise before extracting results.

Common Mistakes

• Using constant material properties. Stiffness, yield strength and expansion coefficient all vary with temperature, and across a large temperature swing that variation matters.
• Assuming a uniform temperature. Skipping the transient thermal step ignores the gradients that often cause the worst stresses.
• Running a purely elastic analysis. If stresses exceed yield, an elastic model overpredicts stress and cannot capture the plastic strain range driving damage.
• Modelling the constraints too simply. Treating a bolted or bonded joint as perfectly fixed or perfectly free can change the answer significantly.
• Reading results from the first cycle only. The response often needs several cycles to stabilise before it represents long-term behaviour.

Related reading: see our articles on FEA for Gears and Drive Components, Large Deformation Analysis for Nonlinear Mechanical Components and FEA in Product Design, explore our FEA Services and CFD Services (for the heat transfer side of coupled problems), or contact our engineering team to discuss a project.

Frequently Asked Questions

What is the difference between thermal fatigue and ordinary mechanical fatigue?

Mechanical fatigue is driven by externally applied cyclic loads, while thermal fatigue is driven by stresses created internally by temperature change - typically when thermal expansion or contraction is restrained by the component's own geometry, by its constraints, or by neighbouring materials that expand by different amounts. The failure mechanism, cracking from repeated cyclic strain, is similar, but the source of the loading is different.

Why is thermal fatigue usually treated as low-cycle fatigue?

Thermal stresses from large temperature swings can be high enough to cause localised plastic strain in each cycle. When plastic strain is involved, failure typically occurs after relatively few cycles compared with high-cycle fatigue, so strain-based methods such as Coffin-Manson are generally used in place of the stress-based S-N approach common for high-cycle problems.

What is CTE mismatch and why does it matter?

CTE stands for coefficient of thermal expansion. When two bonded or connected materials expand by different amounts for the same temperature change, the difference is forced to be absorbed as strain at the interface. Repeated over many thermal cycles, this mismatch strain is a common cause of fatigue cracking in solder joints, coatings, bonded assemblies and multi-material components.

Does a thermal fatigue analysis need to include plasticity?

Often, yes. If the thermal stresses at the critical location exceed the material's yield strength, a purely elastic analysis will overpredict stress and cannot capture the plastic strain range that drives low-cycle fatigue damage. Whether plasticity is needed depends on the loading severity and material, and should be checked rather than assumed.

Conclusion

Thermal fatigue is easy to underestimate precisely because nothing is visibly pushing on the part. The damage comes from the component being unable to expand and contract freely, cycle after cycle, and it concentrates wherever constraint, temperature gradient or material mismatch is greatest. FEA makes that concentration visible by coupling a transient thermal analysis with a structural one, then reading the strain range at the critical location and converting it into a life estimate.

If a component will be heated and cooled repeatedly in service, the useful question isn't only "how hot does it get?" but "what is stopping it from expanding freely, and how many times will that happen?" Simulation is how you answer both before the first crack appears in the field.

---

For more engineering and simulation insights, explore HyperCurve. If this article helped you, please share it with your colleagues.