Removing recalcitrant organic micropollutants (MPs)—tiny, persistent contaminants that resist breakdown and pose risks to both the environment and public health – is a complex challenge in modern water treatment. These pollutants, which can include pharmaceuticals, pesticides, and industrial chemicals, are often found in water supplies and can be difficult to remove using traditional treatment methods.
Granular activated carbon (GAC) works by trapping micropollutants in its highly porous structure, but the effectiveness of these filters diminishes over time as the GAC becomes saturated and eventually, MPs may begin passing through the filter. Predicting exactly when this breakthrough occurs has traditionally required costly, time-consuming experiments.
This is where a recent article, co-authored by Carollo Engineer’s Yoko Koyama and published in Environmental Science & Technology, comes in. Titled “Machine learning models to predict early breakthrough of recalcitrant organic micropollutants in granular activated carbon adsorbers,” the study offers a groundbreaking solution to this challenge by using machine learning to predict GAC filter performance.
The research team built a database of over 400 data points from GAC studies and applied machine learning techniques to develop models that can predict how long GAC filters will be effective. These models focus on key variables, such as the properties of the micropollutants and the quality of the water, to estimate when breakthrough will occur. Their findings showed that the machine learning model could accurately predict GAC performance without the need for extensive physical testing.
This work not only saves time and money but also empowers water utilities to make faster, more informed decisions about the best GAC products to use in different scenarios. The research represents an important step forward in achieving cleaner, safer water for all.
To read the full article, follow this link.