Share this article
Continue reading with these related posts
Fatigue failures do not announce themselves. A shaft carries its rated load for months, then fractures at a load it has handled thousands of times before. The failure looks sudden. The process was not — it began as a microscopic crack at a stress concentration, growing incrementally with each cycle until the remaining section could no longer carry the load.
An FMEA that lives in a spreadsheet and never changes a design decision is not failure analysis — it is documentation theatre. A useful FMEA is a living tool that identifies the failure modes most likely to cause harm, ranks them by risk, and drives specific design or process changes to reduce that risk before the product reaches the customer. The difference between the two is in how the analysis is structured and what is done with the results.
An overload fracture tells a fast story. There are no beach marks, no gradual progression, no evidence of incremental growth. The fracture happened quickly, often in a single load cycle, and the surface records the load level, the material condition, and whether the failure was expected or anomalous. Reading that surface correctly distinguishes a component that was simply overloaded from one that was brittle when it should have been ductile.
Temperature does not have to reach the melting point to cause failure. The mechanisms through which elevated temperature degrades engineering components are more subtle and often more insidious than outright melting: a sustained temperature a fraction of the way to melting causes creep, where material deforms slowly under loads that would be perfectly safe at room temperature. A temperature that cycles repeatedly across a range causes thermal fatigue, where differential expansion and contraction accumulate damage just as mechanical cyclic stress does. A brief temperature excursion above a material's tempering temperature reduces its hardness permanently, without any obvious visual evidence at room temperature.
These thermal failure mechanisms are frequently misdiagnosed as mechanical failures because the symptoms — dimensional change, cracking, reduced strength — look similar at first inspection. The difference is in the pattern of damage, the microstructural evidence, and the operating condition that caused it. Identifying the thermal mechanism correctly is essential because the corrective action is entirely different from the response to a mechanical failure.
For most engineering metals, the mechanical properties specified at room temperature are reliable up to a threshold temperature — roughly 30 to 40 percent of the melting point on an absolute scale (Kelvin). Below this threshold, the material's yield strength, ultimate tensile strength, and fatigue properties are essentially independent of temperature. Above this threshold, thermal activation becomes significant: atoms have enough thermal energy to move, diffuse, and rearrange, and the mechanisms of creep, oxidation, and phase transformation begin to operate.
For carbon steels with a melting point around 1500°C (1773 K), this threshold falls at approximately 430–530°C. Aluminium alloys, with a much lower melting point of around 660°C (933 K), begin to experience significant creep effects above approximately 150–200°C — temperatures easily reached in engine compartments, exhaust systems, and industrial process equipment. Nickel superalloys, developed specifically for high-temperature service, maintain their properties up to 70 percent of their melting point.
Homologous Temperature Governs Thermal Effects
The relevant temperature for assessing thermal failure risk is not the absolute temperature but the homologous temperature — the ratio of operating temperature to melting point, both in Kelvin. A steel at 500°C and an aluminium alloy at 150°C have similar homologous temperatures (~0.43) and will exhibit similar degrees of creep susceptibility relative to their strength. Comparing absolute temperatures across different alloy systems is misleading.
Overheating occurs when a component is exposed to a temperature above its rated service limit for a period sufficient to cause microstructural change. The damage may be permanent even if the component returns to normal operating temperature and shows no obvious distress.
For steels, the most common overheating effects are:
Over-tempering or anneal of heat-treated parts. Quenched-and-tempered steels are heat-treated to a specific hardness and strength by tempering at a defined temperature after quenching. Exposure to temperatures above the original tempering temperature further softens the material, permanently reducing hardness and strength. This damage is invisible without hardness measurement — the part looks normal but is below specification.
Grain boundary oxidation. At temperatures above approximately 700°C for carbon steels, oxygen diffuses along grain boundaries and oxidises the metal preferentially there. On the fracture surface of an overheated part, this produces a dark, intergranular fracture path rather than the normal transgranular path, visible at 50× or higher under optical microscopy.
Phase transformation. For steels, heating above the austenitising temperature (around 723°C for eutectoid composition, lower for lower-carbon steels) followed by slow cooling in service can produce unintended phase transformations — pearlite or bainite replacing the intended tempered martensite, with different and usually lower mechanical properties.
Recognition of overheating: surface discolouration (blue-to-purple colours on steel indicate temperatures of 280–320°C; straw-to-gold colours indicate 220–280°C; black oxide indicates temperatures above approximately 400°C). Hardness below the specified minimum without visible damage. In severe cases, grain coarsening visible in cross-section metallography, or oxidised surface layer.
Creep is time-dependent plastic deformation under sustained load at elevated temperature. Unlike room-temperature plastic deformation, which stops when the applied stress falls below yield strength, creep continues indefinitely under stresses well below yield — as long as the temperature is above the creep threshold and the load is applied.
The creep curve has three stages. Primary creep occurs immediately after load application: the creep rate is initially high and decreases as the material work-hardens. Secondary creep is the steady-state stage: the creep rate is approximately constant, and this is the stage that governs most engineering life predictions. Tertiary creep is the accelerating stage: creep rate increases as the cross-section reduces and internal damage accumulates, leading to fracture.
For design purposes, the creep rupture stress (the stress that causes fracture after a specified time — 100,000 hours is common for power generation equipment) and the creep strain rate in secondary creep are the governing properties.
Recognition: dimensional change without fracture — shaft elongation, bearing housing distortion, turbine blade lengthening, bolt relaxation. In more advanced cases, creep cracking at grain boundaries (cavitation damage, visible in metallographic cross-section as discrete voids along grain boundaries) and eventual creep rupture with an intergranular fracture path. The fracture surface of a creep rupture shows wedge-shaped voids at grain boundaries and a rough, intergranular morphology without the smooth void-coalescence dimples of ductile overload.
Bolt Relaxation in High-Temperature Joints Is Creep
Bolted flanges and joints in high-temperature service lose clamping force over time not from mechanical vibration or joint settling but from creep of the bolt and joint materials under sustained tensile stress at temperature. This is why high-temperature flange bolts require re-torquing after the first thermal cycle and periodic inspection thereafter. Ignoring relaxation in pressure-containing joints at elevated temperature is a significant safety risk — a flange that was properly torqued at ambient temperature may have lost 30 to 50 percent of its clamping force after one year in high-temperature service.
Thermal fatigue is damage accumulation from repeated thermal cycling. When a component is heated and cooled cyclically, differential expansion between constrained regions generates thermal stresses — tensile during cooling, compressive during heating, or vice versa depending on geometry and constraint. These stresses cycle with the thermal cycle, and their accumulated damage is analogous to mechanical fatigue.
Thermal fatigue cracks typically initiate at surface stress concentrations — corners, notches, holes, and weld toes — and propagate inward perpendicular to the direction of maximum thermal stress. Unlike mechanical fatigue cracks, thermal fatigue cracks are often multiple and may appear in a network pattern (crazing or craze cracking) because the biaxial thermal stress state can initiate cracks in multiple directions simultaneously.
The crack morphology under optical microscopy distinguishes thermal fatigue from mechanical fatigue: thermal fatigue cracks are often filled with oxide (because they open and close at high temperature, allowing oxidation of the crack walls), have a branched, irregular path, and often terminate at grain boundaries. Mechanical fatigue cracks are cleaner, more regular in path, and have less oxide fill.
Common thermal fatigue sites: die casting dies (cycled between molten metal temperature and cooling temperature), exhaust manifolds and headers (cycled between ambient and exhaust gas temperature), furnace components, and any structure with repeated startup and shutdown cycles.
Recognition: network of surface cracks at geometrically stressed locations, oxide-filled cracks, cracking localised to the hottest zones or the zones with greatest thermal gradient. The crack pattern is often described as "crazing" for the fine-scale network, or "elephant hide" for the characteristic appearance of severely crazed surfaces.
Slow Thermal Cycles Are More Damaging Than Fast Ones
Counterintuitively, slow thermal cycles (startup/shutdown over hours) often cause more thermal fatigue damage than fast cycles (brief temperature spikes) because the slow cycle allows more thermal gradient to develop through the section thickness. The thermal stress is proportional to the temperature gradient, not the absolute temperature change. Equipment that is cycled slowly and infrequently may accumulate thermal fatigue damage faster per cycle than equipment that cycles rapidly but stays near thermal equilibrium.
At high temperatures, metals react with oxygen in the environment to form surface oxide scales. Controlled oxidation is beneficial — the oxide scale on stainless steels and nickel alloys acts as a diffusion barrier that slows further oxidation. Uncontrolled oxidation is damaging — it progressively removes material from the surface and can spall, exposing fresh metal to continued attack.
Hot corrosion is an accelerated form of oxidation that occurs when sulphur compounds, vanadium compounds, or chlorides deposit on the metal surface at high temperature and flux the protective oxide scale. In gas turbines, hot corrosion from fuel impurities or sea salt ingestion can remove millimetres of blade material in hours.
Recognition: loose, porous, or spalling oxide scale on the surface; orange-to-black scale depending on temperature and alloy; scale composition visible by energy-dispersive X-ray analysis if a scanning electron microscope is available; metal thickness reduction beneath the scale. In hot corrosion, the scale is more voluminous and less adherent than in normal oxidation.
A carbon steel pressure vessel flange in a chemical processing plant showed persistent bolt loosening requiring re-torquing at every maintenance interval despite being correctly torqued at initial assembly. The operating temperature was 380°C, which was within the rated service temperature of the flange material (A193 B7 bolting).
The problem was not the flange — it was the bolt material. Grade B7 bolting (chromium-molybdenum steel, 4140/4142 type) is rated for high-temperature service, but its creep properties at 380°C were not the primary concern. The concern was the flange gasket: a spiral-wound stainless/PTFE gasket was specified. PTFE has a continuous service limit of approximately 260°C and a maximum short-term limit of approximately 290°C. At 380°C, the PTFE filler was creeping under the clamping load, and the joint was relaxing due to gasket creep, not bolt creep.
The visual evidence was a gasket that had thinned uniformly by approximately 0.6 mm (measured at removal) from its original 4.5 mm thickness — consistent with sustained compression at a temperature well above the material's creep threshold.
Corrective action: replace the spiral-wound PTFE gasket with a spiral-wound stainless/graphite gasket (continuous service limit 650°C) or a metallic ring-type joint gasket (RTJ) suitable for the service temperature. After replacement, no further bolt loosening was observed over the following eighteen months.
Root cause: gasket material specification that did not account for the actual operating temperature. The PTFE gasket specification came from a lower-temperature design revision and was not updated when the operating temperature was revised upward.
Verify homologous temperature before applying room-temperature design properties. For any application above ambient temperature, check whether the operating temperature is a significant fraction of the material's melting point. If the homologous temperature exceeds 0.4 to 0.5, creep and oxidation data are required — not just room-temperature mechanical properties.
Use creep-rated materials and specifications for elevated temperature service. Standard structural steel specifications do not include creep rupture data. High-temperature pressure vessel and piping design standards (ASME Section VIII, ASME B31.3) specify materials with published creep data for the applicable temperature range. Do not use generic material specifications for elevated-temperature pressure-containing applications.
Design thermal cycling range into the thermal fatigue life estimate. For equipment with known startup/shutdown cycles, estimate the thermal stress from the temperature differential and use it in a thermal fatigue life calculation. Reduce stress concentrations at hot locations — generous radii at corners, elimination of abrupt section changes — using the same principles as mechanical fatigue prevention.
Control thermal gradients during startup and shutdown. Thermal fatigue damage is proportional to thermal gradient, not absolute temperature. Controlled startup rates (temperature ramp rates specified in the operating procedure) reduce the transient thermal gradient and extend the life of components susceptible to thermal fatigue.
Thermal Failures Are Often Prevented at the Specification Stage
The most common cause of thermal failure is a specification that does not account for the actual operating temperature — a material chosen for its room-temperature properties, a gasket specified for a lower-temperature design revision, a coating rated for ambient service used in a thermal cycling environment. Systematic review of material and component specifications against the actual temperature range, including excursions and transients, catches most potential thermal failures before the equipment is built.
Thermal failures add temperature as a dimension to the failure analysis space. With the physical failure mechanisms now covered — fatigue, corrosion, wear, overload, and thermal failure — the series turns from diagnosis to prediction. The next post introduces FMEA: failure modes and effects analysis, a structured method for predicting which failures are most likely to occur before they happen, so that design changes can prevent them proactively rather than reactively.
