Bayes Optimal Wastewater Classification Using Noisy Sensors

At the US Armys bases access to freshwater is vital, yet it often comes with a high price tag.Water supply and disposal poses a logistical challenge, and often involves costly trucking.To avoid such costs, the US Army would like to recycle as much wastewater as possible at its bases, including its smaller and more remote operations. For this reason, the US Army is interested in a wastewater classification device (WWCD) that can assist in proper wastewater management and reuse. We propose to develop a real-time WWCD that would automatically classify both treated and untreated wastewater to assist the US Army in meeting its wastewater management objectives.The sensors we propose to use will be field-deployable, robust, and inexpensive.However, they will also be inherently noisy, especially when compared to standard laboratory measurements.Our WWCD will conduct innovative point-of-device analytics to make accurate classifications, despite being given very noisy data.We will use a novel combination of techniques in machine learning, optimal decision theory, and principles of wastewater treatment to develop a classification algorithm for our WWCD, which can provide results in real-time and for a fraction of the costs associated with laboratory methods.