The Dual-Mechanism Failure Mode That Limits Rotor Life

Creep-fatigue interaction (CFI) is the primary life-limiting failure mechanism in steam turbine rotors precisely because it is not a single process but the dangerous convergence of two. During steady-state high-temperature operation—steam temperatures reaching 620–650°C in ultra-supercritical units—time-dependent creep deformation and intergranular void formation accumulate at grain boundaries within the rotor material. During every transient startup and shutdown, cyclic thermal gradients impose tensile and compressive stress cycles at the rotor surface and bore, initiating and advancing transgranular low-cycle fatigue (LCF) cracks.

The interaction is not merely additive. A Chinese patent on steam turbine rotor creep-fatigue life prediction (Huaihe Electric Power Company, Fengta Power Branch, 2021) describes the mechanism precisely: when transgranular fatigue cracks meet grain boundary voids generated by creep, “fatigue cracks and creep voids mutually promote and develop each other, forming a fatigue-creep interaction” that accelerates crack growth beyond what either mechanism would produce alone. This mutual acceleration is the mechanistic foundation for all validation approaches discussed in this article.

Large coal-fired units in the 200–1000 MW class represent the core application domain for CFI validation. Published work explicitly examines high-pressure (HP) and intermediate-pressure (IP) rotors of the K-200-130 turbine class, thermal stress and fatigue life calculations for 1000 MW ultra-supercritical rotors under cold-start conditions, and thermomechanical analysis using unified phase mixture constitutive models that account for rate-dependent inelasticity, hardening, and softening through an iso-strain approach with soft and hard microstructural constituents. Nuclear steam turbine applications are increasingly prominent in the most recent filings, with ASME-III Division 5 frameworks being adapted for molten salt and sodium-cooled reactors.

Real-Time Stress Monitoring: Closing the Loop Between Operation and Damage

Real-time stress monitoring systems validate creep-fatigue interaction damage by instrumenting the rotor or stator with sensors—thermocouples, speed sensors, pressure transducers—and using embedded computational models to track accumulated damage continuously, enabling operators to intervene before critical thresholds are reached. This approach has the longest lineage in the dataset: Westinghouse Electric Corporation filed a rotor surface strain fatigue accumulation supervisory system as early as 1970, computing rotor surface strain and accumulated fatigue damage for operator display with a potential closed-loop strain limit control.

Hitachi’s 1981 patents extended this by determining thermal and centrifugal stresses in real time and comparing them against brittle fracture toughness criteria during cold startup—a critical safety gate given that cold-start thermal gradients produce the highest transient stress severity. The operating logic is directly relevant to modern practice: rotor thermal stress is computed by numerical solution of the heat propagation partial differential equation, with thermocouple measurements from the stator providing boundary input. A 2015 literature study on rotor thermal stress monitoring confirms this approach remains the industry standard input method for embedded lifing systems.

“With the increased use of renewable power there is an increased need for electric network operation to operate with increased cycling… increased exposure to frequent thermal transients increase the risk of the occurrence of thermal fatigue crack initiation during cold, warm and hot start-ups as well as during shutdowns.”

GE Infrastructure Technology’s control system for managing steam turbine rotor stress, filed in 2017, specifically addresses combined-cycle steam turbines being cycled due to renewable integration. It manages rotor thermal stress accumulation across cold, warm, and hot startups—distinguishing between three distinct thermal transient severity levels whose cumulative damage contributions differ substantially. GE’s wireless creep life management system (2016) adds a further dimension: first portions attached to rotating components transmit response signals representative of rotatable component measurements to a processor for continuous life tracking, removing the need for physical slip rings or telemetry contacts.

For nuclear steam turbine applications, the Shanghai Power Equipment Research Institute’s 2023 filing tracks stress corrosion crack growth life, LCF crack growth life, and high-cycle fatigue crack growth life in parallel—running three concurrent life consumption models whose outputs are aggregated for comprehensive crack propagation life safety monitoring. This multi-track architecture reflects the more stringent regulatory requirements governing nuclear components under standards bodies such as IAEA.

Computational Damage Models: From Miner’s Rule to Nonlinear Interaction

Computational life prediction methods validate creep-fatigue interaction by calculating a fatigue damage fraction (from cyclic strain ranges using approaches such as Coffin-Manson) and a creep damage fraction (from time-at-stress using creep constants), then combining them against a damage criterion boundary—either the ASME-NH creep-fatigue interaction diagram or a nonlinear equivalent. Three distinct combination strategies appear in the dataset, spanning linear, ductility exhaustion, and nonlinear fitting approaches.

Linear and time-fraction accumulation

The ASME-NH framework provides the standard platform: elastic analysis methods with creep-fatigue interaction rules and specified damage limits are applied to elevated-temperature cyclic service components. According to published ASME-NH design studies for process reactor components, engineers extract local stress, maximum elastic strain range, and temperature at the critical location, then compute fatigue and creep damage fractions that must together satisfy the interaction diagram. This framework is now being extended by ASME Division 5 to nuclear-cladded components for molten salt and sodium-cooled reactor service.

Ductility exhaustion

Solar Turbines’ condition-based lifing patents (US, WO, 2015) explicitly employ a ductility exhaustion method to combine fatigue and creep strain rates per load cycle. A fatigue module, creep module, and ductility exhaustion module work in series: the fatigue module calculates cycle-by-cycle fatigue strain increments, the creep module calculates time-integrated creep strain increments during hold periods, and the ductility exhaustion module sums both against the material’s available ductility to determine remaining useful life. According to this dataset, ductility exhaustion is the dominant damage combination method in active patent filings.

Nonlinear interaction parameter fitting

Jiangsu Fangtian Electric Power Technology’s 2025 patent moves beyond linear accumulation. It classifies operating conditions into five distinct modes—startup, steady-state, peak load ramping, peak load reduction, and shutdown—and calculates per-phase creep damage Dc and fatigue damage Df using mode-specific creep constants and Coffin-Manson parameters. A nonlinear three-point curve-fitting function then determines the interaction parameter governing the creep-fatigue damage trajectory, finding the intersection with the damage criterion boundary to calculate remaining life and formulate maintenance strategies. This operating-condition segmentation is a materially more granular approach than the simple start-stop counting methods that characterised earlier filings.

Literature supports the model diversity: the strain range partitioning approach for a steam turbine rotor steel uses finite element strain range partitioning to model LCF life, while the multi-surrogate collaboration approach (MSCA, 2020) combines a dynamic neural network surrogate (DNNS) with a distributed collaborative strategy to quantify uncertain creep-fatigue damage—computing a reliable life figure of 629 cycles at 0.998 7 reliability for the studied turbine rotor. Standards organisations including ASME continue to refine code provisions as these probabilistic methods mature.

NDE-Based Crack Sizing and Remaining Life Projection

Non-destructive evaluation (NDE) based remaining life assessment validates creep-fatigue interaction damage by detecting and sizing internal flaws through ultrasonic testing, then projecting remaining life by calculating how detected crack sizes will grow through combined creep crack growth and fatigue crack growth over future loading blocks. This methodology—established by Westinghouse in 1992 and refined by Siemens Energy between 2012 and 2014—remains strategically significant because it grounds the life assessment in measured physical evidence rather than model assumptions alone.

Westinghouse’s foundational patent established the core methodology: assign critical crack sizes based on fracture toughness, accumulate creep crack growth and fatigue crack growth per loading block, and decrement backwards to find the allowable initial indication size. In practice, this means ultrasonic indications from routine outage inspections are directly linked to a remaining cycle count, enabling engineering-judgment-based decisions on continued service, monitoring interval, or repair. According to the patent dataset, the Westinghouse 1992–1993 and Siemens Energy 2012–2014 filings anchoring this cluster have since lapsed or become inactive—creating potential open space for novel NDE-to-remaining-life pipeline approaches.

Siemens Energy’s deterministic fatigue life prediction system (US/WO/EP/CA, 2012–2014) refined the approach by incorporating flaw interaction rules: when ultrasound data identifies adjacent or overlapping indications, the system merges them into an effective initial crack size before applying crack growth models. This is a materially important step because unresolved adjacent flaws that independently appear sub-critical may be collectively above the critical threshold. The calculation then proceeds to determine the number of load cycles until the critical crack size is reached, explicitly accounting for both fatigue and creep growth contributions per cycle.

Physical coupon testing as complementary validation

Where NDE detects existing flaws, physical coupon or witness specimen methods validate the creep behaviour of rotor materials without destructively testing the rotor itself. Alstom (Switzerland) developed the primary coupon-based methodology (US, 2002 and 2003): test elements are mounted at operating loads comparable to the monitored component, exposed for predetermined operating periods, then removed and examined to derive the component’s creep behaviour from the test element’s post-exposure state. A parallel approach described in literature uses structurally similar elements cut from finished parts for LCF testing, combined with three-dimensional FEM analysis of the specimens to estimate full-size component strength. According to ISO and international testing standards, physical specimen correlation remains a required validation step for new material qualifications in high-temperature rotating equipment.

Patent Landscape Shifts: China’s Emergence and the ASME-NH Standard

The geographic structure of CFI validation innovation has changed substantially since 2020. In this patent dataset spanning 1970–2025, the dominant jurisdiction by historical filing volume is the United States, anchored by General Electric / GE Infrastructure Technology (at least 8 filings, 2009–2021), Solar Turbines (at least 5 filings, 2015–2022), Siemens Energy (4 filings, 2012–2017), and Westinghouse Electric (2 filings, 1992–1993). European assignees include Rolls-Royce PLC, Alstom (Switzerland), GE Technology GmbH, and Arabelle Solutions France.

The structural shift since 2020 is the emergence of China as the most active filing jurisdiction in the 2021–2025 window. Chinese assignees in this dataset include Jiangsu Fangtian Electric Power Technology (CN, 2025, CFI evaluation method for thermal power units), China Nuclear Power Engineering (CN, 2025, creep-fatigue assessment for nuclear equipment), Shanghai Power Equipment Research Institute (CN, 2023, nuclear steam turbine rotor monitoring), Huaihe Electric Power / Fengta Power Branch (CN, 2021, steam turbine rotor creep-fatigue life prediction), and Xi’an Jiaotong University (CN, 2024–2026, random spectrum embedding fatigue reliability analysis). This concentration of 5+ recent filings signals a rapidly maturing domestic capability for both fossil and nuclear steam turbine CFI validation.

Mitsubishi Hitachi Power Systems represents a distinct materials-design approach: multiple filings in CA and US jurisdictions (2015–2017) address 12Cr steel with 900–1200 MPa tensile strength, targeting the balance between stress corrosion cracking (SCC) susceptibility and LCF life simultaneously. This is the only major assignee cluster in the dataset that addresses CFI damage mitigation primarily through material selection rather than monitoring or modelling—a complementary strategy with implications for IP freedom-to-operate analysis.

From a standards perspective, the ASME-NH framework is becoming the de facto code platform for both fossil and nuclear CFI validation. R&D teams entering this space should architect their computational tools to produce the local stress, maximum elastic strain range, and temperature inputs that these code frameworks require—whether the target is power plant design approval or regulatory compliance. The extension of this framework into ASME-III Division 5 for nuclear cladded components is an active standards evolution tracked by organisations including the US Nuclear Regulatory Commission.

Emerging Directions: Probabilistic Methods and Renewable-Driven Cycling

Two structural forces are reshaping CFI validation methodology: the mathematical shift from deterministic to probabilistic life assessment, and the operational shift from baseload to cycling duty driven by renewable integration. Both forces converge to make existing damage accumulation models—designed for steady, predictable operating histories—insufficient for the actual loading spectra that modern power plant rotors experience.

Probabilistic and surrogate-model reliability assessment

Xi’an Jiaotong University’s random spectrum embedding method for blisk fatigue reliability (CN, 2024–2026) applies probabilistic distribution fitting to measured parameter data, representing an emerging probabilistic dimension to what has historically been deterministic life assessment. The multi-surrogate collaboration approach (MSCA) in the literature (2020) demonstrates the scale of departure from deterministic methods: it combines a dynamic neural network surrogate with a distributed collaborative strategy, quantifying uncertain creep-fatigue damage and computing a reliable life figure at a specified reliability degree rather than a single point estimate. The result—629 cycles at 0.998 7 reliability for the studied turbine rotor—provides quantified confidence bounds that deterministic calculations cannot offer and that asset managers increasingly require for risk-informed inspection interval decisions.

Renewable integration and flexible operation demands

The growth of wind and solar generation has forced previously baseload steam turbines into cycling duty, dramatically increasing startup and shutdown frequency and thereby the rate of CFI damage accumulation per calendar year. GE Technology GmbH’s steam turbine rotor patent (EP, 2016) explicitly acknowledges this operational reality, and research at Doosan Skoda Power on the influence of hold periods on residual fatigue life confirms the engineering response: every additional start-stop cycle contributes a discrete increment of fatigue damage, while the hold period at operating temperature contributes a time-integral of creep damage whose severity depends on temperature, stress level, and hold duration.

The strategic implication for product developers is clear: condition-based lifing systems that can distinguish cold-start, warm-start, and hot-start thermal transient severity—mapping each startup type to its specific cumulative damage fraction—are commercially differentiated offerings for utilities facing grid-driven cycling mandates. Jiangsu Fangtian’s five-mode operating condition segmentation (startup, steady-state, peak load ramping, peak load reduction, shutdown) is the most granular operational decomposition in the current patent dataset, and it represents a model for how future systems will need to characterise highly variable load histories.

Across all four validation approaches—real-time monitoring, computational damage models, NDE-based crack sizing, and physical coupon testing—the common trajectory is toward greater integration: monitoring systems that feed computational models, computational models that define NDE inspection triggers, and NDE results that update model parameters in a closed-loop architecture. The China Nuclear Power Engineering 2025 filing, with its real-time ASME-NH total damage formula output to the main control room alarm system, is the most complete realisation of this integrated architecture in the current dataset. Tracking how this architecture evolves—particularly in the Chinese domestic nuclear and ultra-supercritical fossil fleets—will be a leading indicator of where CFI validation methodology moves next across global energy infrastructure.