Data science for metrology

It is a common misconception that the automation of data transmissions is done automatically. However, many factors, such as weather conditions, the functioning of the communication chain, the calibration of sensors, storms, floods, etc. can disrupt a measurement. We regularly observe sensor failures and drifts, bad settings, but also transmission anomalies.

In a rapidly changing digital environment, the operational performance of teams depends on reliable and qualitative data. Whether in drinking water production plants, distribution networks, sewage networks or waste treatment plants, the monitoring of key indicators is essential to maintain activities. For example, a drinking water production plant must constantly monitor the quality indicators of the water it serves. 

With traditional supervision, analysis is done on an ad hoc basis by the operator consulting the raw data displayed by the supervision software. This monitoring is then done manually and punctually, not allowing to respond to crises quickly or even to anticipate them. Since the sensors are treated as a whole, none of them are prioritized within the large panel deployed in the field. Product manager Martin Vuillaume and the data scientists at SUEZ Smart Solutions therefore came up with a monitoring solution that would go further: "Unlike standard monitoring, our offer allows us to focus our efforts on the most critical and potentially faulty sensors, to characterize suspected malfunctions remotely, and to anticipate possible deviations or behaviors that could lead to irreversible breakdowns."

Let's take the example of this graph which represents the analysis of a turbidity measurement (water quality parameter) over 3 months. As you can see on this picture, the measurement drifts very slowly, which is very difficult to detect with a so-called classical supervision, and checks made with the naked eye once a week or once a day.

Smart Metrology Services was able to detect this drift automatically and the experts understood that the problem could be due to a clogged sensor: an intervention to remedy this could be programmed.

Thanks to several years of experience in data science for the environmental sector, SUEZ Smart Solutions has created algorithms that automatically detect sensor-specific anomalies.

Thus, Smart Metrology Services has about fifteen "off-the-shelf" algorithms that can be associated with each analyzed measurement (noise, min/max, atypical data, regularity, multiple correlation, etc.). From the results of the algorithms, a synthetic score is calculated to characterize the status of each measurement.

Thanks to this continuous analysis of sensor data and the automatic detection of suspected malfunctions, operators can make significant gains in terms of savings and operational performance. The number of field interventions can be optimized, and equipment requires less maintenance than if certain problems had worsened over time.

Smart Metrology Services is a solution that is available now to anyone who needs reliable data. With the growing demand for the deployment of specialized metrology sensors, SUEZ is optimistic about the future of this offer and is already deploying it within its own plants to improve team performance!