August 2021 Monitoring water quality is a first important step to provide water security for any drinking water provider. An event detection system (EDS) allows you to catch possibly dangerous events in real-time.
There is a variety of s::can products that make event detection easy and accessible.
Our software ana::tool (e.g. as part of a micro::station) turns your monitoring system into an event detection system using spectrometry for unambiguous detection of contaminations and increased information through wavelength patterns.
It identifies unknown and unusual conditions and enables operators to react timely to faults in the monitored water system, determines normality of these data, and triggers an alarm when a significant deviation from normality is detected. This allows you to take immediate counter measures and to prevent harmful effects.
ana::tool is simple, easy to use and automatic. Its unmatched event detection tools are based on proven algorithms for real-time event detection that use data streams from all connected probes separately and in combination. The software is optimized for use of multi-dimensional spectral data, but will also work with single or multiple one-dimensional inputs.
The blue line in this graphic shows measurement data from lab conditions that was spiked with simulated contaminations of different substances. The baseline is perfectly flat with a little bit of measurement noise and we can clearly see the spikes of the simulated events.
The green line shows the same contaminations to a background from a field measurement, the situation changes dramatically. The contaminations signals sit on top of large baseline variations and becomes a challenging task for an EDS to distinguish between events and natural fluctuations.
Only with a full optical UV-Vis spectrum such events can be detected with utilizing the s::can spectral alarm, as it completely separates events from the background. (This would be impossible with single wavelength UV254 sensors)
The pattern alarm makes use of a set of reference data which is regularly updated during training processes. This reference data represents the normal, regular water quality.
Incoming measurements are compared to these references and the distance to the reference data is used as alarm value. If the alarm value exceeds a predefined threshold, the pattern alarm indicates a water quality event.
Event Detection by pattern recognition is more powerful than a than threshold alarm, it detects small deviations that are not visible to single parameter alarm systems. It also monitors the normal range of inputs as well as correlations between inputs.
The figure above shows the UV absorption at 254nm at the inlet (orange) and at the outlet (green) of a drinking water plant. Usually, natural fluctuations of the organic load of the raw water are efficiently removed by the water works. However, around the 1st of November an increase in organics was passing through the plant which was also measured at the outlet with some time delay.
This sudden increase was also picked up by the event detection system ana::tool as can be seen on the dataset shown below and the plant could react immediately. In this case the full spectrum measured by the spectro::lyser at the outlet of a plant was used for the event detection – the so called spectral alarm.