Real-world case studies

District heating
The first of five real-world case studies applies uncertainty propagation, correlation analysis and sensor fusion aspects as well as distributed network techniques to assess heat flow and temperature sensors for district heating. The current focus is on applying the methods for analysis of correlated sensor data, physical modelling, and other information (e.g. historical data) to create an initial topology of the network. Meter data is being used to find static and dynamic correlations between the different meters to reduce temperature measurement uncertainty.

High value heat treatment
The second real-world case study applies uncertainty propagation, correlation analysis and sensor fusion aspects to improve process temperature control in high-value heat treatment applications. The current focus is on the investigation of a multi-wire temperature sensor (thermocouple), whereby the systematically different calibration drift of each wire in the thermocouple enables a combinatorial approach to applying a correction. Techniques to make use of data from multiple distinct sensors in a particular process to deduce calibration drift are also being developed.

Natural gas and hydrogen distribution networks
The third real world case study applies sensor fusion aspects and distributed sensor networks techniques to improve uncertainty of flow metering in natural gas and hydrogen distribution networks. It is now evident that a large proportion (around 50 %) of gas flow meters are subjected to insufficiently frequent re-calibration. A survey among industry experts and stakeholders is facilitating a root cause analysis for the appearance of calibration drift in gas flow meters; this will feed into developments of techniques for detecting the need for re-calibration of flow meters in the networks.

Air quality monitoring
The fourth real world case study will exploit mixed air quality sensor network self-calibration and co-calibration methods (uncertainty-aware sensor fusion, drift detection and Laplace-domain tools) with emphasis on the dynamic behaviour of the sensor networks under non-static conditions. A literature survey of wireless network calibration strategies in air quality monitoring for smart cities is being prepared, and improved calibration strategies are being developed.

Building monitoring applications
The fifth real-world case study applies automatic calibration of mechanistic models and the impact it has on the accuracy of sensor networks in building monitoring applications, as well as the extent to which adding soft sensors can increase control performance. Modelica or Matlab models will be trained using the framework developed, and, using data quality metrics and traceability information, the models will be applied to representative buildings which are part of the real-world use-case demonstration.