Article

How to Run Better Slope Stability Assessments in Slide3

Published on: Mar 14, 2025 Updated on: Mar 14, 2025 Read: 5 minutes

The accuracy of your 3D slope stability assessments depends on how well your limit equilibrium method identifies the lowest factor of safety. Some approaches require too many adjustments, while others may overlook key failure surfaces or get trapped in local minima and lead to costly design errors.

In a webinar hosted in 2024, Dr. Sina Javankhoshdel discussed Intelligent Search in Slide3, a feature that improves how failure surfaces are identified and optimized. Here are the key takeaways from the session to help you run more reliable analyses:

Not All Search Methods Are Created Equal

In slope stability analysis, search methods define how slip surfaces are generated, and they should find the lowest factor of safety without unnecessary computation to improve speed. Two advanced search methods, Particle Swarm Optimization (PSO) and cuckoo search, which are good at finding critical slip surfaces, can fail to identify the complex slopes like those with weak layers or deep-seated failures. They cannot reliably be used alone to identify the most critical failure surface.

To improve accuracy, Intelligent Search builds upon existing methods and focuses on high-risk areas first, rather than distributing slip surfaces randomly. This reduces computational inefficiencies and makes sure that critical failure surfaces are not overlooked. Also, for the most reliable results, global optimization methods for search should be combined with local optimization techniques to refine the identified failure surfaces.

Monte Carlo Optimization Isn’t Always Enough

Monte Carlo is one of the traditional methods used to locally optimize slip surfaces, however it can get stuck in local minima, meaning it could optimize the wrong failure surface and give a false sense of stability. The biggest limitation of Monte Carlo is that it assumes the critical slip surface is already a good surface, which can lead to inaccurate results for initial failure surfaces that are not well-defined.

In one example from the webinar (see Figure 1 below), Monte Carlo failed to detect a critical weak layer deep in the soil profile, producing a higher factor of safety. Engineers should never rely on Monte Carlo alone, since this is an outdated method and less reliable compared to modern methods.

For the same model, Surface Altering Optimization (SAO), a tool for local optimization of the surfaces found by cuckoo search and PSO, was able to find the critical slip surface that was going through the deep weak layer in the model.

Local Optimization is Crucial in 3D

Unlike 2D models, where a well-chosen search method can often find an accurate failure surface, 3D models require additional local optimization to refine results. SAO is a must-use tool for this purpose, according to Dr. Javankhoshdel, as it is more accurate and significantly faster than Monte Carlo.

In 3D slope stability models, disabling SAO almost always results in an overestimated factor of safety. If a slip surface is not being refined properly, its shape may be too simplified to capture the true failure mechanism. The lesson is clear: in any 3D slope stability analysis, turning on Surface-Altering Optimization will provide more reliable results.

Figure 1. An example of a weak layer plane, where the left model used SAO and has a FoS of 1.018, while the right model only used Monte Carlo and has an FoS of 1.104.

Why More Search Exploration Improves Accuracy

3D models have a much larger solution space than 2D models, meaning they require more search explorations to properly identify failure surfaces. Engineers often underestimate how many searches are needed for convergence to an accurate factor of safety.

Increasing searches in PSO and cuckoo search allows the model to search deeper before settling on a failure surface. However, these methods rely on random sampling and cannot guarantee full model coverage. They are also prone to getting trapped in local failures, which can result in missing the true critical slip surface.

For engineers working on large-scale slope models, the best practice is to either gradually increase search iterations until the results stabilize or use a more thorough method like Intelligent Search.

Intelligent Search: A Smarter Starting Point

Intelligent Search optimizes 3D slope stability analysis by identifying high-risk regions first, rather than distributing slip surfaces randomly. The Region of Interest (ROI) feature allows you to visualize failure-prone areas before running a full analysis.

Figure 2. An example of how the Intelligent Search process functions.

Key benefits of Intelligent Search include:

Automatically identifying failure-prone regions before running a full search.
Focusing computational power on high-risk areas to reduce time spent on calculations.
Finding more localized slip surfaces using the Regions of Interest (ROI) feature, which visually highlights likely failure regions.
Ensuring multiple failure modes are captured, rather than just the first slip surface found.

By refining the search before computation starts, Intelligent Search eliminates guesswork and reduces the need for manual trial-and-error adjustments.

Figure 3. A model that had Intelligent Search applied; note the low value FoS areas that were thoroughly captured. “Using Intelligent Search, every slip surface is going to be there.” – Dr. Sina Javankhoshdel

Final Thoughts

To improve 3D slope stability analysis, you can:

Avoid relying on Monte Carlo optimization alone, as it assumes the slip surface is already a good surface.
Always enable Surface-Altering Optimization (SAO) to refine failure surfaces.
Increase search explorations for complex projects to avoid overestimating the factor of safety, or use Intelligent Search to focus on high-risk areas from the start.

By applying these best practices in your geotechnical risk analysis, you can conduct better slope stability assessments, avoid critical mistakes in 3D slope modelling and create more reliable designs.