Wind turbines parked in high winds don’t simply stand still — their blades can still flex, vibrate, and experience aerodynamic loads that engineers need to predict accurately. A peer-reviewed study published in Frontiers in Energy Research in March 2026 by Bangga, Syawitri, and Chairunnisa put the tools used for those predictions to a rigorous test — and found significant gaps.
The researchers assessed how well existing dynamic stall models capture aerodynamic loads on a wind turbine airfoil across four stall-induced vibration regimes, including conditions where airflow strikes the blade in reverse — opposite to its normal operating orientation.
Study tests four stall-induced vibration regimes across full angle-of-attack range
The study focused on the S814 airfoil, a 24%-thick profile commonly used in wind turbine blades. Researchers ran OpenFOAM CFD simulations across the full angle-of-attack range — from −180° to +180° — capturing aerodynamic behavior under every flow orientation a parked or idling turbine might encounter.
Four stall-induced vibration (SIV) regimes were tested: SIV-1 covers normal flow conditions, split into positive and negative angle-of-attack ranges, while SIV-2 covers reverse flow conditions, again split into positive and negative ranges. Identical reduced frequency and amplitude settings were applied across all four regimes to allow direct comparison.
CFD results were then compared against three semi-empirical engineering models — Øye, Beddoes–Leishman (BL), and IAG — all run within BLADED, the industry-standard wind turbine design software. The comparison targeted conditions specifically relevant to turbines parked under extreme winds or exposed to significant yaw misalignment, where accurate load prediction is critical for structural safety.
Reverse flow aerodynamics driven by asymmetric vortex structures that models cannot replicate
In normal operation, air flows from a blade’s leading edge toward its trailing edge. Reverse flow — the SIV-2 regime — flips that orientation entirely. Incoming air strikes the airfoil from the trailing edge side, fundamentally changing how pressure builds, where separation begins, and how vortices form and shed.
The CFD simulations captured this complexity in detail, revealing early flow separation, irregular vortex shedding, and wide aerodynamic hysteresis loops. These features make reverse flow conditions considerably harder to model than normal flow.
A pronounced asymmetry also emerged between the positive and negative angle-of-attack ranges within SIV-2. The study attributes this to the airfoil’s camber and geometry, which cause directionally dependent vortex evolution when flow is reversed — behavior that wouldn’t appear in a symmetric airfoil profile.
Perhaps most significantly, conventional definitions of stall onset do not apply in SIV-2. The airfoil effectively behaves as if stalled across nearly its entire operating range under reverse flow — a finding that undermines core assumptions built into standard engineering models.
IAG model outperforms Øye and Beddoes–Leishman models but still underestimates peak loads
In the normal flow regime, the three engineering models performed very differently. The IAG model captured hysteresis loops with reasonable accuracy, while the Øye and BL models significantly underestimated hysteresis width, missing a key feature of dynamic stall behavior.
In reverse flow, the gap between CFD and all engineering models widened considerably. All three produced over-smoothed predictions and failed to reproduce the complex, vortex-driven load variations that the OpenFOAM simulations resolved.
Quantitative error analysis confirmed IAG’s relative advantage. Across all four SIV regimes, it achieved the lowest L2-norm error — roughly half the error of both the Øye and BL models. Even so, the IAG model fell short of capturing the full amplitude of unsteady loads, particularly in SIV-2.
One adjustment improved IAG’s SIV-2 performance: setting the critical stall normal force to zero, better reflecting the near-constant stall state of the airfoil under reverse flow. Strong unsteady load amplitudes still remained underrepresented, pointing to deeper structural limitations in current semi-empirical frameworks.
Findings carry direct implications for structural safety of large parked wind turbines
Standstill operation is not a safe harbor for wind turbine blades. When a turbine is parked under extreme wind or significant yaw misalignment, blades can still experience high ultimate loads and dynamic instability. The study situates its findings squarely within this known engineering challenge.
The stakes are rising as turbines grow larger. For machines rated above 15 MW, blade lengths increase substantially, and inaccurate load predictions carry more severe consequences. Field data cited in the study suggest that long blades can become unstable in reverse flow at wind speeds as low as 8 m/s — well below the approximately 40 m/s design extreme wind speed specified in standard load cases. If engineering models underestimate loads in SIV-2, fatigue calculations and structural assessments for large blades could be systematically optimistic.
Input data quality also emerged as a critical variable. When experimental wind tunnel measurements were used as model inputs instead of CFD-derived or extrapolated polar data, engineering model predictions improved noticeably. The authors ranked input data quality from best to worst as: experimental measurements, high-fidelity 3D CFD, 2D CFD results, and extrapolated data. Polar extrapolation — currently the most common industry approach for SIV-2 — sits at the bottom of that ranking.
Authors call for new modeling frameworks and further research into reverse flow physics
The study’s conclusions are direct: reverse flow dynamic stall remains one of the least understood aerodynamic regimes in wind turbine research, and current tools aren’t adequate for the task.
Future engineering models will need to treat normal and reverse flow as fundamentally distinct physical problems — developing separate time-delay and flow-memory representations for each regime rather than applying a single framework across the full angle-of-attack range. The authors also recommend identifying alternative physical markers for stall evolution in SIV-2, moving away from conventional stall onset definitions that proved inadequate for reverse flow conditions.
Because the asymmetry observed between positive and negative SIV-2 ranges is tied to airfoil camber and trailing-edge geometry, results may differ substantially for symmetric profiles. Broader testing across airfoil geometries is needed before general conclusions can be drawn.
Key takeaways
The study establishes several clear findings for researchers and engineers working on wind turbine load assessment. CFD simulations using OpenFOAM captured reverse flow aerodynamics with the greatest fidelity, revealing hysteresis loops and vortex dynamics that no current engineering model could fully reproduce. Among the three semi-empirical models tested, IAG consistently produced the lowest errors across all four SIV regimes — but still underestimated peak unsteady loads in reverse flow.
All three models — Øye, BL, and IAG — failed to replicate the complex, asymmetric vortex behavior that defines SIV-2. Standard stall onset definitions don’t hold in reverse flow, and adapting them through simple parameter adjustments offers only partial improvement.
For the wind industry, the practical implications are significant. As turbines grow larger and blade instability under standstill conditions becomes a more pressing concern, the gap between what current models predict and what actually occurs in reverse flow could translate into real structural risk. Closing that gap will require new modeling approaches, better input data, and continued investigation into the physics of reverse flow aerodynamics.
If you want to learn more about this discovery, the full study is available here: Bangga G, Syawitri TP and Chairunnisa (2026) Dynamic stall modeling of a wind turbine airfoil at various stall-induced vibration zones. Front. Energy Res. 14:1761590. doi: 10.3389/fenrg.2026.1761590
Carlos is an engineer with strong expertise in technical and industrial topics. He previously worked at international companies such as Siemens and speaks Spanish, German, English, and Italian.
