The need for accurate sensing of physical quantities such as temperatures, pressures, flows, concentrations and compositions is critical in many monitoring and control applications. The range of available physical sensors and measuring instrumentation is vast, but nonetheless there are several challenges facing the market, the most apparent being:
- Physical sensor systems are costly and require frequent calibration and maintenance.
- Physical sensor systems are often not usable in confined or harsh operational environments (e.g. due to limited physical space, extreme temperatures or pressures, contamination, etc.).
- Increasingly stringent government regulations regarding emission reduction, monitoring and control require new solutions to overcome technical barriers.
Given these challenges, virtual sensing techniques, also known as inferential, soft, or proxy sensing, are software-based techniques (alternative to traditional hardware sensors) aimed at providing feasible and economical alternatives to costly or unpractical physical measurement devices and sensor systems.
In practice, the purpose of a virtual sensing system is to estimate an unmeasured quantity of interest by exploiting the correlated information available from other sensors and measurements. The EEVS technology developed at IFE is based on a novel combination of empirical neural network modelling with ensemble modelling to provide more accurate, more robust and more stable virtual sensor estimations.
The EEVS technology will be applied to two cases to test its applicability. The first case is performed in cooperation with the Norwegian Institute for Air Research (NILU) and concerns air quality monitoring. For the second case, it is suggested to apply the technology for monitoring the oil-in-water concentrations in produced water discharged from offshore oil-separators.
At present, three patents are pending approval, one related to the EEVS technology in general and two related to specific applications of the EEVS technology.