Development of an in-situ detector for classification and regression of dissolved gases in liquid waste with application to wastewater monitoring

Monitoring volatile compounds in sewer systems is of high importance due to the toxic and corrosive nature of various nuisance chemicals generated such as hydrogen sulfide (H2S). Hotspot monitoring facilitates identification of the location of the generated H2S, and thereby targeted treatment can be applied which eventually minimizes the use of chemicals and lowers the environmental effect within the sewer system. Here, we developed a portable detector that automatically extracts and delivers sewer contents to a microfluidic-based detector, fabricated by a selective microchannel embedded with a metal oxide semiconductor (MOS) sensor. Using a wide concentration range of H2S and ammonia (NH3) dissolved in water (i.e., two components to which the MOS sensor has potential cross-selectivity), a database for a machine learning model was developed. The model could classify between NH3 and H2S with 96.4% and 96.9% overall recall in separate and mixture aqueous solutions, respectively. Overall regression precisions of 84.6% and 88.8% were obtained in separate and mixture aqueous solutions, respectively. The developed setup was used in a field test (at Annacis Island (Delta, BC)) wastewater treatment plant where the results showed that the device could identify H2S and NH3 in raw influent samples and measuring the concentrations via regression with 94.6% and 83.5% overall recall and precision for H2S and NH3, respectively. These results demonstrate the promise of the developed automated detector and machine-learning data processing methodology for applications in in-situ wastewater monitoring or treatment through the detection of H2S hotspots for targeted mitigation efforts.