Auto-Refine Circular Search
Automatically refining grid-and-radius search for critical circular surfaces.
The auto-refine circular search is the default method. It evaluates a grid of circle centers and radii, then iteratively concentrates the search around the best-performing region to home in on the critical circular surface — without you having to specify a grid by hand.
How it works
- The slope is divided into stations along its length.
- For each pair of stations, a set of trial circles is generated (each pair of surface entry/exit points spans a circle family).
- Every trial circle is evaluated for its factor of safety.
- The best-performing circles are retained, and a refined set of circles is generated around them.
- Steps 3 and 4 repeat for the specified number of iterations, progressively narrowing the search around the most critical region.
Candidate circles are evaluated in parallel across a pool of background workers, so the search uses available CPU cores.
Parameters
| Parameter | Unit | Default | Description |
|---|---|---|---|
| Divisions along slope | — | 10 | Number of stations along the slope used to generate circle entry/exit points. More divisions give finer spatial coverage. |
| Circles per division pair | — | 10 | Number of trial circles generated for each pair of stations. |
| Iterations | — | 6 | Number of refinement passes. Each pass narrows the search around the retained best circles. |
| Retained percent | % | 50 | Percentage of the best circles kept and refined between iterations. |
These values are also adjusted together by the search complexity slider; the defaults above correspond to the medium setting.
When to use circular search
Circular surfaces are appropriate for:
- Homogeneous or simply layered slopes where failure is rotational.
- A fast first pass to locate the general failure region before refining.
- Cases where a circular assumption is acceptable per the design standard.
If the critical mechanism is controlled by a weak seam, a foundation contact, or other planar feature, a non-circular method (block search or a metaheuristic) is usually more representative. You can also follow a circular search with surface-altering optimization to let the critical circle deform into a lower-FS non-circular shape.