Parametric sweeps

Parametric studies sweep one or more axes across an OpenFAST run and group the results so you can plot trends, build a power curve, or scan a sensitivity envelope. They sit between Single Simulation and Certification: more cases than a smoke test, but you pick the cases instead of inheriting the IEC DLC matrix.

When to use parametric vs certification

• Use parametric when you control the axes: a power-curve build (wind speed only), a sensitivity to turbulence intensity, a structural scan over blade stiffness, an operational scan over yaw misalignment.
• Use certification when the campaign is driven by IEC 61400-1 — DLC matrix, partial safety factors, PDF deliverable. See Certification.

Creating a parametric study

From the dashboard click New Study, pick Parametric Analysis, name it, and hit Start Parametric Analysis. You land in the workbench with the step sequence Turbine → Environment → Parameter Sweep → Run Sweep → Results.

The axis model

The Parameter Sweep tab groups available axes into four categories. Each axis appears as a card you enable, expand, and configure with either a uniform (min, max, step) grid or an explicit list of values.

• Environmental — wind speed, turbulence intensity, shear exponent, seed count. Use these for power curves and IEC-style fatigue scans.
• Operational — blade pitch, rotor speed, yaw angle, cut-in / cut-out wind speed. Use these to test controller tweaks or off-design operation.
• Structural — tower height, blade density scale, blade stiffness scale. Use these for design-space exploration on a custom turbine.
• Aerodynamic — rotor diameter, blade twist scale, blade chord scale. Use these for upscaling / downscaling experiments.

Combination modes

When you enable more than one axis, the sweep editor asks how to combine them. Three modes:

• Factorial — full cross-product. Three wind speeds × five TIs × six seeds = 90 cases. Use this when you actually want the matrix.
• One-at-a-time (OAT) — sweep each axis from its baseline independently. The total is the sum of axis lengths minus duplicates of the baseline. Cheap, gives you per-axis sensitivities, misses interaction effects.
• Paired — pair each axis position-by-position (axis 1 value i with axis 2 value i). Length must match. Use this for “at each wind speed, use a specific TI” pairings that come from a wind climate.

Cost sanity-check before submit

The sweep editor shows the resolved case count and the estimated token cost in real time as you tune axes. The platform enforces a 500-job hard cap per sweep; matrices above that need either a tighter axis range or OAT instead of factorial. Reasons to keep the count low:

• OpenFAST run time is roughly linear in n_cases; large sweeps tie up the slow lane for hours.
• Token cost scales the same way; the cost panel makes the hit obvious before submit.
• You usually learn the shape of a sensitivity from 20–40 cases. A 500-case sweep is rarely the right answer on the first pass — narrow down, then widen.

Submitting and monitoring

Move to the Run Sweep tab. The summary table lists every case with its axis values; click any row to see the resolved OpenFAST inputs it would write. Hit Run Parametric Sweep and the batch goes to simulations:slow. The Run tab swaps into a live progress view: completed / running / queued / failed, per-case status, and a sim-time bar on the active jobs.

Refresh-safe — close the tab and the batch keeps running. The global Job Monitor page (sidebar) shows the same batch alongside any others your org has in flight.

Post-processing

Once a sweep completes, the Results tab grows two parametric- specific surfaces alongside the per-case time-series view:

• X-Y plot — pick an axis for X (e.g. wind speed), a channel reduction for Y (mean, max, std, DEL on the post-transient window), and overlay multiple channels. The classic power-curve view.
• Sweep matrix — a sortable table of every case with the chosen channel reduction. Export to CSV for downstream notebooks.

Letting Zeph build the sweep

Most parametric setups are easier to describe than to click through. “Power curve on the IEA-15 onshore: wind 4 to 24 m/s in 2 m/s steps, three seeds, turbulence IEC-B” is one Zeph turn and an Apply card. See Zeph, your AI assistant.

Tips

• Name your sweeps with the axis in the title (e.g. iea15-ws-sweep-IEC-B) — you'll have a dozen of these in the studies list within a week.
• Pin the random seed when scanning operational axes.Otherwise you can't tell whether the trend you see is from yaw angle or from turbulence sampling.
• Start with OAT, escalate to factorialonce the per-axis effect is understood. Half the parametric studies you'll ever run don't need a factorial.

Next steps

Use Results & analysis for the channel reference, or hand the same workflow to a script via the API.