The Internet of Things (IoT) is ushering a paradigm change in the concept of sensors. Humans are no longer the only users of data. More often than not, data generated by sensors is further used by automatic algorithms, which imposes higher standards for the actionability and interpretability of the data. Attaining these standards typically requires mathematical modeling, which is used to tie together multiple data sources and to learn the relevant information. This leads us to the concept of a virtual sensor.
A virtual sensor is a system that determines, in real time, some physical variable. However, instead of measuring it directly like a regular sensor, a virtual sensor combines modeling with data from possibly several sources. The quantity of interest does not even need to be directly observable, but may be inferred from indirect measurements only. Virtual sensors are becoming increasingly widely used, and in this blog, we explore some of the main reasons behind their popularity and discuss some applications.
Virtual sensors can achieve signal quality beyond what is possible with traditional sensors. For example, an inertial navigational system uses GPS, motion sensors (accelerometers), and rotation sensors (gyroscopes) to determine the position and orientation of an object accurately. Without the motion sensors, the rotation sensors cannot accurately give the tilt angle and the GPS alone cannot track small changes in position. Only the full combination of sensors yields a stable result.
Virtual sensors can also achieve a very low price point. As a simple example, given a person’s age, gender, and weight, the calories burnt during exercise correlate with their pulse. Hence, an expensive metabolic measurement system can be replaced with a pulse meter and some modeling to estimate the calorie burn rate during exercise. Similarly, in some cases, a single expensive sensor can be replaced with several cheap alternatives. A virtual sensor can merge these measurements, in the spirit of sensor fusion, to give a higher quality measurement at a lower price.
Some interesting quantities, such as the physical boil-off rate of liquefied natural gas (LNG) in an LNG tank—the rate at which liquid evaporates to gas—are simply unmeasurable in the traditional sense. In such cases, a virtual sensor is the only possible approach. To this end, Eniram has developed a virtual sensor that takes various thermodynamic measurements together with LNG flow measurements to reconstruct the boil-off rate inside tanks.
Virtual sensors can be used to fix the known shortcomings of other sensors. For example, onboard speed through water (STW) logs often suffer from calibration errors, which could ruin a performance assessment of the vessel and make comparisons between vessels unfair. To overcome this pitfall, Eniram has developed a virtual STW log, which makes STW-related data from different vessels comparable. In this case, the virtual sensor makes use of the onboard STW log data, but is able to estimate STW more accurately by also using other data sources (such as propeller RPM and torque, speed over ground, and current forecasts).
Stay tuned for another blog on this subject. Do you have comments or questions on this topic? Feel free to contact us!