Case Study: AI for Contamination Plume Classification
Artificial intelligence helps diagnose contamination plume growth.
Background
International Environmental Firm engage Remi AI to develop artificial intelligence methods for contamination plume classification.
Determining whether a contamination plume is shrinking or expanding is a key metric in site remediation projects. This is often conducted with a array of monitoring bores around the likely perimeter of the underground contamination and then using a variety of sampling techniques to determine concentration and changes over time.
These samples are then used in their modelling to determine directional movement of the contamination and the binary classification of whether the plume is shrinking or expanding. This modelling is time consuming and an expensive part of the process.
Remi AI worked with the client to deploy our AutoML methods. These are founded on our own Neural Architecture Search methods, where there is meta-a.i searching for the most accurate a.i models for the given task.
We worked with the client to train them in how to use these methods on a range of datasets, but with the key focus on historic contamination plume datasets.
You can see more about our AutoML approaches here: https://www.remistudios.io/
Approach
94% accuracy of contamination plume classification.
Results
After over 180 hours of artificial intelligence searching, our AutoML approach found a a.i classification model that achieved 94% accuracy on historical datasets. This is currently being tested on new jobs, helping Environmental Scientists identify plume change quicker.