An online monitoring system based on an array of non-specific sensors was used for the detection of chemical pollutants in wastewater and water. By superimposing sensor profiles for defined sampling window, the identification of data points outside these normal sensor response patterns was used to represent potential pollution episodes or other abnormalities within the process stream. Principle component analysis supported the detection of outliers or rapid changes in the sensor responses as an indicator of chemical pollutants. A model based on the comparison of sensor relative responses to a moving average for a defined sample window was tested for detecting and identifying sudden changes in the online data over a 6-month period. These results show the technical advantages of using a non-specific based monitoring system that can respond to a range of chemical species, due to broad selectivity of the sensor compositions. The findings demonstrate how this non-invasive technique could be further developed to provide early warning systems for application at the inlet of wastewater treatment plants.