Wind 
Giant wind turbines go haywire when the wind hits from the wrong direction and no one can fully explain what happens next
When a massive wind turbine shuts down during an extreme storm, its blades don’t simply rest. Parked and exposed to howling winds from an unexpected direction, the blades can experience violent, chaotic aerodynamic forces — forces that engineers struggle to measure, model, or predict with any confidence.
The problem starts with geometry. Wind turbines are designed for flow entering at the leading edge of the blade. But under certain standstill conditions, the wind arrives from the opposite end — the trailing edge. That reversal puts the blade in aerodynamic territory that existing engineering models were never built to handle.
When a parked turbine is anything but at rest
A turbine that isn’t spinning might seem safe. But standstill operation — when a turbine is parked or idling during extreme winds or severe yaw misalignment — can expose blades to some of the highest loads they ever experience. Without rotation to distribute forces evenly, the blade absorbs whatever the wind delivers.
The key mechanism is stall-induced vibration, or SIV — a feedback loop between blade movement and aerodynamic separation that amplifies rather than dampens itself.
High ultimate forces and dangerous vibrations can develop even with the rotor stationary. The key mechanism is stall-induced vibration, or SIV — a feedback loop between blade movement and aerodynamic separation that amplifies rather than dampens itself.
The problem compounds as turbines scale up. Field data suggest that blades on turbines rated above 15 MW can become unstable in reverse flow at wind speeds as low as 8 m/s — a moderate breeze, not an exceptional storm. Next-generation turbine scale is pushing designs into territory engineers didn’t originally anticipate.
Normal stall versus reverse flow stall: a fundamentally different beast
To understand why reverse flow is so difficult to model, consider dynamic stall under normal conditions. When a blade pitches rapidly past its static stall limit, flow separation doesn’t happen instantly — a leading-edge vortex builds, then sheds, creating pronounced loops in lift and drag. Engineers call this hysteresis.
Researchers classify SIV into two regimes: SIV-1 covers normal flow, where air travels from the leading edge to the trailing edge, and SIV-2 covers reverse flow, where air arrives from the trailing edge side.
In SIV-2, adverse pressure gradients develop earlier, separation bubbles form differently, and vortex shedding follows patterns with no analog in normal flow. The airfoil effectively behaves as stalled across nearly its entire operating range, and forces fluctuate in ways existing models weren’t designed to track.
Putting engineering models to the test
A recent study published in Frontiers in Energy Research quantified how badly current models fail. Researchers used OpenFOAM to simulate airfoil behavior across the full 360-degree angle-of-attack range, then compared results against three widely used semi-empirical models — Øye, Beddoes–Leishman, and IAG — run inside BLADED, the industry-standard design software.
The test airfoil was the S814, a 24%-thick profile common in wind turbines. Validation used wind tunnel data from Ohio State University. One key numerical finding: capturing vortex dynamics accurately required 3,000 time steps per oscillation cycle — coarser resolutions obscured the critical features entirely.
Asymmetry, vortices, and the limits of existing models
The CFD results revealed a pronounced asymmetry between positive and negative SIV-2. Positive SIV-2 produces larger, less coherent vortical structures and irregular hysteresis loops. Negative SIV-2 yields a more repeatable response — but still highly unsteady.
Against those results, the engineering models struggled. Øye and Beddoes–Leishman produced over-smoothed predictions that missed vortex-driven features entirely. The IAG model performed considerably better — its L2-norm errors were roughly half those of the other two models across all four SIV regimes. Even so, IAG significantly underestimated hysteresis amplitudes in SIV-2. No current model captures the full picture.
What drives the chaos — and what must change
The parametric analysis identified a clear hierarchy. Reduced frequency — how fast the blade oscillates — has a far stronger effect on reverse-flow hysteresis loops than Reynolds number does. Higher oscillation frequencies produce fuller, more energetic loops, while Reynolds number effects remained comparatively modest even when doubled.
The asymmetry between positive and negative SIV-2 traces back to airfoil geometry. Camber and trailing-edge shape drive different vortex behaviors depending on flow direction — meaning the effect is specific to real-world blade profiles, not a generic aerodynamic curiosity.
The study also flags a practical data problem: experimental wind tunnel measurements outperform CFD-generated data, which in turn outperforms the polar extrapolation methods the industry most commonly relies on for SIV-2 design cases. Future models must treat reverse flow as a distinct physical regime, with its own stall-onset definitions, time-delay parameters, and flow-memory representations.
What comes next
The findings point toward several concrete needs: better physical markers for stall evolution in reverse flow, distinct formulations for SIV-2 conditions, and higher-quality experimental input data rather than extrapolated curves never validated in this regime.
As turbines grow larger, standstill aerodynamics will only become more consequential. Blades that go unstable at 8 m/s in reverse flow aren’t an edge case — they’re a design challenge affecting every parked turbine in a storm. The tools engineers currently use were built for a different problem. Building models that can actually navigate it is the work ahead.
If you want to learn more about these findings, check the full study 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
Author Profile
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.
