Introduction

The following overview and summary table explain the general differences between three automatic search algorithms available in GeoStudio: Particle Swarm Optimization, Differential Evolution, and Cuckoo Search. The comparison focuses on search behaviour, convergence characteristics, and computational performance.

Particle Swarm Optimization (PSO): Explores the search domain using a population of candidate solutions that respond to their own best positions and the best positions found by the swarm. It may converge quickly as the swarm concentrates around promising solutions, although this can limit further exploration of other potential failure mechanisms.

Differential Evolution (DE): Generates trial solutions using differences between population members, followed by crossover and selection. These mechanisms can maintain population diversity and support continued exploration, sometimes resulting in more gradual convergence.

Cuckoo Search (CS): Uses Lévy flights to introduce occasional large jumps in the search space, supporting exploration beyond the current promising region. Its computational performance varies with the problem; it may require more time in some cases and converge faster in others.

 

Important Note

Each algorithm may perform best under different conditions, depending on geometry, material properties, groundwater conditions, loading, search settings, and model complexity. No algorithm is universally the fastest or guaranteed to identify the global minimum factor of safety (FOS).

Summary Table

Algorithm	Global Exploration	Local-Minimum Considerations	Computational Performance
Particle Swarm Optimization	Explores through individual and shared knowledge within the swarm.	Early convergence of the swarm may concentrate the search around a local minimum.	Can converge rapidly, but performance depends on the problem and settings.
Differential Evolution	Explores through population differences, mutation, and crossover.	Population diversity can support exploration of alternative solutions, but local convergence remains possible.	May converge more gradually; no fixed speed ranking applies.
Cuckoo Search	Combines smaller movements with occasional larger jumps through Lévy flights.	Larger jumps can help explore beyond a local minimum, but do not guarantee the global minimum.	May be faster or slower than other algorithms, depending on the problem and settings.

 

Automatic Search and Slip-Surface Optimization

Automatic search explores candidate slip surfaces within the defined search domain. Critical slip-surface optimization subsequently adjusts an identified surface to seek a lower FOS. Although some algorithms can be used in both processes, their role depends on the selected analysis settings.

When Random Walk is selected for optimization in SLOPE/W, it performs local refinement by randomly adjusting points along an existing slip surface and progressively refining its segmentation. Its results can be sensitive to the starting surface and optimization settings. Local optimization should therefore complement, rather than replace, adequate initial searching.

 

Practical Considerations

Compare suitable algorithms using consistent modelling assumptions and assess whether increased search effort materially changes the minimum FOS or failure mechanism. Review the resulting geometry against the geology and expected failure behaviour, as a lower FOS alone does not establish physical realism.