Sensor Failure Detection in Water Quality Monitoring

Abstract

in the treatment of raw water, to make it suitable for use as drinking water, it is necessary to monitor and control key parameters such as colour, pH and turbidity. Signal condition monitoring is essential when the automatic control of one or more parameters is dependent on the quality of the measurements made. Operational failure of process plant instrumentation can lead to reduced production and plant shutdown. This paper considers a number of possible methods of identifying faulty sensors including multiple regression, and artificial neural networks, and also methods of estimating values to allow for continued plant operation.