Wastewater Treatment & Biofilms
Biofilms and AI in the wastewater treatment process at the School of Physics and Astronomy
There is a particular type of wastewater treatment called the Activated Sludge Process (ASP), which is an area of particular interest for me. In the ASP, wastewater goes from households and industry to a wastewater treatment facility, where the water is treated in several stages before being returned to the water cycle.
A crucial stage in wastewater treatment is the ‘Biological Treatment Process’. Here, a complex plethora of microorganisms, mainly bacteria, degrade the dissolved organic waste. For this to happen, the bacteria must be able to replicate, grow and divide, and then eventually the bacteria must stick together to form large dense clumps that can be easily separated from the water. The water then goes onto the next stage for further treatement. Crucial to this process is the presence of filamentous bacteria, too many or too few and the process breaks down. This can be very costly to the treatment process and the operators then must retrospectively add environmentally harmful and expensive chemicals to alleviate the problem.
What we’re doing here at the University of Edinburgh with Veolia is trying to better understand the biophysical mechanisms by which these filamentous cells mediate the aggregation of the bacteria in the wastewater treatment process.
There is both healthy and unhealthy sludge, and in order for operators to assess the health of their sludge within their wastewater treatment plants, they take regular samples from the tanks in which these bacteria reside. They then image the samples under a microscope to determine if the sludge is healthy or not, which can be very subjective. Our research with Veolia is attempting to integrate machine learning algorithms into this process to automate the decision making, so that ideally, we can take a microscopy image from a sample and then the algorithm can determine the health of the sludge. Veolia have provided us with thousands of images from their sites across the world to help better train our machine learning algorithms to modernise the wastewater treatment industry.
Another issue we have is with the filamentous cells. When these cells become too abundant, we have a problem called ‘filamentous sludge bulking’ and we hope that our machine learning algorithm will support us in developing a sludge characterization platform. Through this platform, we would aim to revolutionize and modernize protocols used to identify the onset of filamentous sludge bulking, which would make it easier for operators to assess the risk of it occurring in their system so they could take preventative action instead of retrospectively adding the environmentally harmful and expensive chemicals.