Project Leads:
Dr Luis Bandeira Neto, Experienced Engineer
Dr Michael Munro, Principal Engineer
Team Members: Darren Pemberton (Project Design Reviewer), Nathan Podlich (Code and Implementation Reviewer), Alan English (UX Reviewer).
Â
Luis, how did the innovation project idea originate?
The idea came from a project I was working on that required running multiple repetitive GeoStudio analyses. The models were large and slow, and basic data extraction was very time-consuming. When I discovered that GeoStudio has an API, I realised that the process could be automated with Python and run in the background with only minor user input.
What problem does your project solve?
The framework significantly reduces the time required for repetitive GeoStudio analysis tasks. The code basically reduces one day of work to 10 – 12 minutes. Processes that used to take days or weeks can now be completed in a much shorter timeframe, and manual changes that once took around 30 minutes now take seconds. The software can be utilised outside of working hours, leaving a series of simulations to be run after hours.
While this code was initially developed for a specific use case, the same API-driven approach can be applied to other repeatable tasks across GeoStudio workflows.
Can you explain your approach?
The original concept focused on automating a seepage model in GeoStudio by using Python to test different input values, assess the results and identify an optimal solution. Although that specific model was no longer being used, GeoStudio is still utilised across ATC Williams, so the approach had realistic potential for other applications.
The idea was then adapted to a closure capping scenario for a Tailings Storage Facility (TSF), where cover material needs to be placed and assessed over the long term. This provided a practical case study to understand the GeoStudio API better, test its scalability and explore how the same automation framework could be applied to more complex tasks involving multiple variables, such as layer thickness, equipment and material options.
Were there any challenges or unexpected discoveries?
One unexpected development was that the project evolved from a Python automation task into an AI-assisted workflow. After discussing a similar task with Michael Munro (Principal Engineer), he applied AI to take the concept further and automate more of the process. While this is a promising outcome, the results need to be reviewed carefully to confirm they are reliable and safe to use. Because I have studied the API and understand how the framework functions, I will be able to review the code and ensure the procedure is being performed as intended.
What stage is the tool at now, and what is planned in the future?
The core automation framework has been partially coded and is close to achieving the intended workflow. The next stage is to use the remaining project budget to review the AI-assisted components of the code Michael has created. This review strengthens the reliability of the code’s outputs, determines best practices and identifies any required refinements. Ultimately, it will ensure that the tool can be safely applied to future Closure Capping design tasks.
Michael, how has this approach been applied to projects?
Building on Luis’s concept, AI-assisted coding was used to extend the framework into a capping assessment tool that drives SLOPE/W end-to-end through the GeoStudio API. The tool takes the measured tailings strength profiles for a site and automatically works through the equipment and capping options such as machine, layer thickness and geotextile. It finds the lightest plant and thinnest capping layer that can safely cap each part of a TSF beach and then determines how far that capping can safely extend toward weaker ground.
It has been used for tailings capping assessments for projects at Mangoola and West Wallsend, where batches of over a hundred SLOPE/W analyses run unattended (a recent run solved 117 models in about 23 minutes) and produce results as factors of safety, capping-limit plans and client-ready figures. This is work that would previously have taken weeks of manual modelling.
Can the API be applied to other GeoStudio tasks?
As we build knowledge over the GeoStudio API and its limitations, more can be done on the automation of other tasks undertaken with the software. There are easily identifiable wins for running lengthy sensitivity analysis and optimisation cases, allowing our engineers to spend less time with the software and more analysing the problem itself. Moreover, there are APIs in other software used by ATCW that allow for similar cases to be explored. AI lowers the bar on accessing the code and accelerating implementation, but the base knowledge of the problem fundamentals is more important than ever to ensure the outputs are defensible and reliable.
Learn more about ATCW’s Innovation Projects Initiative