Steam Turbine Rotors Weren't Built for This: A Q&A on Fatigue Life in the Energy Transition

In a previous article, Majid Nazemi, Mechanical Simulation Engineer at ENGIBEX, described a thermo-mechanical FEA workflow for estimating steam turbine rotor fatigue life under flexible operation. In this Q&A, he answers follow-up questions on the real cost of rotor failures, how far simulation models can be trusted, access to OEM data, inspection strategy, regulation, data quality and real-time lifetime monitoring.

1. The energy transition as a game-changer

You describe how the shift to renewable energy is forcing thermal plants to start and stop far more frequently than they were designed for. In your view, are plant operators and OEMs truly prepared for this new reality, or are we collectively underestimating the accumulated damage already happening across the fleet?

The risk of steam turbine rotor cracking due to thermal cycling has been known for many years, but the operating conditions of many thermal power plants have changed significantly with the energy transition. Turbines that were originally designed to operate predominantly under stable baseload conditions are now increasingly required to start, stop and change load more frequently. As cracking indications began to appear across different units, this prompted further investigation into how these more demanding operating profiles were affecting rotor lifetime.

It became clear that repeated thermal transients during startups and shutdowns can generate significant cyclic stresses, particularly at geometric stress concentrations, and accelerate fatigue-life consumption. In response, plant operators have increasingly carried out remaining-life and lifetime-consumption assessments to understand the condition of existing rotors and determine where inspection or operational changes may be required.

OEMs have also responded by improving turbine designs, optimizing critical geometries and developing or selecting materials better suited to flexible operation. So I would say that the industry is now much better prepared and the mechanisms are much better understood. However, the challenge remains particularly important for the existing fleet: many turbines operating today were designed for a different operating regime, and their actual accumulated damage depends strongly on their individual operating history. For these units, we should not simply assume that the original design lifetime remains representative; their actual operating history needs to be considered when assessing their remaining life.

2. The cost of a rotor failure — real numbers

The article mentions "prolonged unavailability and significant loss of power generation" as consequences of rotor failure, but stays general. Can you give a sense of the actual financial stakes? What does a major rotor failure — repair, replacement, lost generation — typically cost an operator in practice?

It is difficult to give a representative number because the financial impact is highly plant- and failure-specific, and in my experience the detailed commercial figures associated with individual cases are generally confidential. The cost also goes far beyond the rotor itself. Depending on the extent of the damage, the operator may face inspection and engineering costs, machining or weld repair, rotor replacement, disassembly and reassembly of the turbine, and potentially damage to other components.

More importantly, there is the cost of unplanned unavailability. A major rotor problem can keep a unit offline for a significant period, and the financial impact of lost generation depends strongly on the unit capacity, electricity prices, contractual obligations and how long the outage lasts. For this reason, I would be cautious about quoting a generic figure. What can be said with confidence is that the economic consequence of a major rotor failure can be substantial, which is why lifetime assessment, targeted inspection and early identification of critical locations are so valuable.

3. Baseload-designed rotors in a flexible world

Many rotors currently in service were designed for near-steady baseload operation. In your assessment, what proportion of the existing fleet is operating significantly outside its original design envelope today — and is that a ticking clock?

I would not want to put a percentage on the existing fleet because I do not have sufficiently representative data to support such a figure. Individual asset owners and fleet managers are in a much better position to know how many of their units were originally designed for predominantly baseload operation and how their operating profiles have changed over time.

What is important is that these turbines are not simply replaced as soon as their operating regime changes. Many can continue operating safely, but their actual operating history needs to be considered. A rotor may still be operating within its allowable operating limits while accumulating fatigue damage faster than originally anticipated because of more frequent startups, shutdowns and load changes.

This is exactly where the lifetime-assessment workflow described in the article becomes valuable. By reconstructing the actual operating history and estimating the resulting fatigue-life consumption, operators can make informed decisions about the next step. Depending on the results and the condition of the component, this could mean continued operation, modifying startup or shutdown procedures to reduce thermal stresses, performing targeted inspections during the next outage, repairing a critical region, introducing additional monitoring, or eventually replacing the rotor.

So rather than describing the existing fleet as a “ticking clock,” I would say that it is a fleet whose condition needs to be understood and managed based on its actual operating history. The key question is not simply how old a rotor is, but how it has been operated and how much of its fatigue life has been consumed.

4. FEA vs. reality — how often does the model get it right?

You are rigorous about the sources of uncertainty in the simulation chain. But looking at cases where FEA predictions have been validated against actual rotor condition (inspections, failures), how well do state-of-the-art thermo-mechanical models actually perform? Where do they consistently fall short?

Validation is a critical step before such a modelling workflow is used for lifetime assessment. One approach is to apply the methodology retrospectively to turbines for which the operating history and inspection results are already known. If cracking has been detected at a particular location after a known operating history, for example, we can examine whether the model identifies the same region as critical and whether the predicted level of fatigue consumption is consistent with what has been observed in the field. The comparison is not perfect, because an inspection tells us when a crack was detected, not necessarily exactly when it initiated.

There are certainly cases where the predictions do not fully match the observed condition of the rotor. The difficulty is that this is a modelling chain: steam conditions, heat-transfer coefficients, transient temperatures, material properties, local stress and strain calculations, fatigue models and damage accumulation all contribute to the final result. When there is a discrepancy, identifying which part of that chain is responsible is not always straightforward.

One area where further improvement is particularly important is the way fatigue damage is calculated from the local stress-strain history. Conventional low-cycle-fatigue approaches can simplify a much more complicated reality. A turbine rotor experiences non-isothermal and history-dependent loading, potentially including plasticity, stress relaxation, mean-stress effects and, at sufficiently high temperatures, interaction between creep and fatigue. More advanced assessment procedures, including approaches based on R5 and more sophisticated constitutive material models, can account for some of these effects.

So I would not describe FEA as giving an exact prediction of when a rotor will crack. Its strength is in combining the physics of the component with its actual operating history to identify critical locations and estimate lifetime consumption. The more the individual parts of that workflow are validated against measurements, material data and actual inspection findings, the more confidence we can have in the resulting lifetime assessment.

5. The OEM knowledge wall

You flag that OEM drawings may be incomplete or protected by IP restrictions, making independent assessments harder. Is this a genuine barrier in practice, and do you think the industry should push for greater transparency around critical safety-relevant geometry data for in-service assets?

Yes, this can be a genuine barrier in practice. For an independent lifetime assessment, the geometry at critical locations needs to be represented accurately because relatively small differences in features such as grooves, fillets or other stress concentrations can significantly affect the calculated local stresses and strains. However, sufficiently detailed OEM drawings are not always available to the asset owner or to the engineering team carrying out the assessment.

At the same time, there is strong competition between turbine OEMs, and detailed design information represents valuable intellectual property. There is therefore an understandable tension between protecting proprietary design knowledge and providing asset owners with enough information to assess the integrity of equipment that may remain in service for several decades.

In practice, one way of overcoming this limitation is through reverse engineering. Techniques such as 3D scanning and silicone replication can be used to reconstruct the as-built geometry of critical locations when sufficiently detailed drawings are unavailable. This allows the required thermo-mechanical assessment to proceed without necessarily requiring access to the complete OEM design documentation.

Whether the industry should move toward greater transparency is a broader question, but from an engineering perspective, what matters is that operators have access to sufficiently accurate information—or reliable means of obtaining it—to assess the condition and remaining life of critical in-service components.

6. Inspection vs. simulation — where should operators invest?

If a plant operator has limited budget, would you recommend investing primarily in better FEA modelling and monitoring, or in more frequent physical inspection? Is there a sweet spot between the two that the industry still hasn't found?

I would not see simulation and physical inspection as competing investments because they provide different types of information. FEA helps us understand where fatigue damage is most likely to accumulate and how the operating history may have consumed the lifetime of the rotor. Importantly, this analysis can generally be performed while the turbine remains in operation, using the available operating history, although developing and validating a detailed model can take considerable engineering effort.

Physical inspection, on the other hand, provides direct information about the actual condition of the component, but it generally requires access to the rotor and therefore has implications for outage planning, time and cost. Continuous monitoring, including vibration monitoring, provides another useful source of information during operation, although it should not be considered a replacement for targeted inspection.

The most effective approach is therefore to combine these tools. FEA and operating data can be used to identify the most critical locations and estimate their lifetime consumption. Monitoring can provide additional information during operation, while physical inspections can then be targeted toward the locations and time periods where they provide the greatest value. If the analysis indicates significant lifetime consumption or an elevated risk, this can justify bringing an inspection forward rather than simply waiting for the next scheduled overhaul.

So, particularly when budgets are limited, I think the objective should be risk-based inspection rather than simply “more simulation” or “more inspection.” The value of the simulation is that it can help operators decide where, when and how urgently physical inspection is needed.

7. Regulatory catch-up

Fatigue life management of turbine rotors relies largely on internal company standards, OEM recommendations, and codes such as EN 13445 or ASME. In your opinion, are current regulatory frameworks adequate for the new flexible operating regime, or is there a risk that regulators are still calibrated for a baseload world?

There is an important distinction between general design codes and the procedures used specifically for turbine rotor life management. Standards such as ASME provide valuable engineering frameworks, but there is not one universal ASME procedure that tells an operator exactly how to assess the remaining fatigue life or fracture risk of an in-service steam turbine rotor under flexible operation. In practice, operators therefore rely on a combination of OEM procedures, industry guidance such as EPRI recommendations, inspection results and engineering assessments.

The good news is that the industry is aware of the problem. Existing guidance already recognizes the importance of cyclic operation and increasingly uses approaches based on the actual operating history of the turbine. This can include reconstructing thermal transients, calculating the resulting stresses and strains using FEA, and estimating fatigue damage or crack development.

So I would not necessarily say that the engineering community is still calibrated for a baseload world. The tools for addressing flexible operation do exist and continue to evolve. The challenge is that they are spread across OEM procedures, industry guidance and different engineering codes rather than being captured in one universally applicable regulatory framework specifically for steam turbine rotor lifetime management.

8. The human factor in plant operations

Your methodology relies on plant instrumentation and operating data. From experience, how often is that data actually clean, complete and trustworthy enough to feed directly into a lifetime assessment — and what happens when it isn't?

That is a very practical challenge. When working with operating histories collected over many years, some degree of data-quality issue is very common. There may be missing periods, sensor spikes, drift, inconsistent measurements or values that are physically unrealistic. For this reason, raw plant data should not simply be fed directly into a lifetime-assessment model.

Data validation and cleaning are therefore an important part of the assessment. Sometimes an unrealistic result from the simulation is actually the first indication that something is wrong with the input data. You then have to trace the result back through the calculation and determine whether the problem comes from a measurement, a missing data point, the reconstruction of the steam conditions, or the model itself.

When unreliable data are identified, they need to be corrected or reconstructed on a defensible engineering basis—for example, using neighboring measurements, redundant instrumentation, known operating conditions or physically consistent interpolation. It is also important to document these corrections because uncertainty in the operating data ultimately contributes to uncertainty in the lifetime estimate.

More advanced data-analytics and anomaly-detection techniques can certainly help automate part of this process, particularly when many years of operating history need to be processed. However, I would still consider engineering judgment essential: the data need to make sense not only statistically, but also physically in terms of how the turbine actually operates.

9. Reduced-order models and the future of real-time monitoring

You mention reduced-order models as a practical compromise for processing long operating histories. Do you see a near-term future where a lightweight digital twin running in the plant's control system provides a live fatigue-damage readout after every startup? What are the main obstacles still standing in the way?

This is not only a future possibility—it is already being implemented. Dashboards can be designed that use calibrated reduced-order or surrogate models in the background to track the evolution of rotor lifetime consumption during operation. Instead of running a full thermo-mechanical FEA after every transient, the reduced-order model can process the operating data much more quickly and provide an updated estimate of fatigue-life consumption, for example after each startup or shutdown.

The important point, however, is that the reduced-order model is only as reliable as the detailed model used to develop and calibrate it. The full FEA workflow is still needed to establish the relationship between operating conditions, rotor temperatures, local stresses and strains, and fatigue damage at the critical locations. The reduced-order model then provides a computationally efficient approximation of that behaviour.

One of the main challenges is therefore ensuring that the reduced-order model remains valid across the range of operating conditions that the turbine actually experiences. If operating conditions change significantly or the turbine experiences transients that were not adequately represented during calibration, the model may need to be checked and recalibrated. Uncertainties in plant measurements, material properties, thermal boundary conditions and the underlying fatigue model also remain—they do not disappear simply because the final calculation is performed in real time.

So the technology for near-real-time lifetime monitoring already exists. The main challenge is not necessarily making the calculation fast enough, but ensuring that the number shown on the dashboard remains a reliable representation of the physical condition and accumulated damage of the actual rotor.

10. Rotor lifetime management as a competitive advantage

Beyond risk reduction, could better fatigue management become a genuine commercial differentiator — allowing plants to operate more aggressively or offer faster grid services while competitors hold back? Or do you think the industry will remain inherently conservative on this?

Yes, I think it can provide a genuine commercial advantage. Better fatigue management is not only about preventing failures; it can also help operators use their assets more efficiently.

If an operator has a good understanding of how different operating transients consume rotor life, they can make better-informed decisions about when and how flexibly to operate the plant. For example, instead of treating every startup in the same way, the operator can understand how different startup procedures affect fatigue consumption and potentially optimize the operation to reduce unnecessary damage while still responding to the needs of the grid.

There is also an important maintenance benefit. If the actual condition and lifetime consumption of the rotor are better understood, inspections and maintenance can be targeted based on risk rather than relying only on fixed intervals or conservative assumptions. This can help reduce unnecessary maintenance and avoid unplanned downtime, ultimately reducing operating and maintenance costs. In that sense, lifetime assessment can become more than an engineering safety tool. It can provide information that supports operational and commercial decisions. The industry will understandably remain conservative when component integrity is concerned, but better modelling does not necessarily mean taking more risk. Ideally, it allows operators to understand and manage that risk more intelligently

Interviewee: Majid Nazemi 

Mechanical Simulation Engineer at Engibex