Qualitative representation of trends (QRT) as a tool for automated data-driven monitoring of on-line sensors

The application of on-line sensors for monitoring or real-time control of wastewater treatment plants remains a challenging problem. The primary cause for this is the unknown quality of the data obtained by such on-line instruments. In the presented work sensor-internal meta-data is used for self-diagnosis. It is verified whether the complete measurement profiles obtained during regular measurement cycles of a phosphate analyzer can help to assess the sensor's status and even indicate potential faults causing faulty behaviour. A qualitative analysis approach is taken and the obtained results indicate that measurement profiles are indeed affected by the implied fault scenarios and may thus serve for proper fault detection and diagnosis.