Monitoring wastewater BOD using a non-specific sensor array

An electronic nose incorporating a non-specific sensor array of 12 conducti ng polymers was evaluated for its ability to monitor wastewater samples. Se wage samples collected from the inlet works, settlement tank and final effl uent outlet over 5 months (January-May) were used to correlate the sensor r esponses of these samples with their corresponding 5-day biochemical oxygen demand (BOD) values. Canonical correlation analysis (a linear reduction te chnique) showed that the relationship between the sensor responses and BOD over the 5 months was non-linear. However, the separate analyses of subsets of these samples for shorter time periods showed that a Linear relationshi p is apparent for time periods of 4 weeks or less, suggesting a correlation window occurs between the sensor responses and BOD. Preliminary neural net work analysis supported these observations and using a three-layer backprop agation network showed that it is possible to predict BOD values from elect ronic nose analysis of a wastewater sample. The findings suggest that elect ronic nose technology could be used for the noninvasive monitoring and/or c ontrol of a wastewater treatment process. (C) 1999 Society of Chemical Indu stry.