Virtual Sensors
A virtual sensor is a what the name implies.
Instead of having a real sensor transmitting data about some physcial property of a system,
data from another, unrelated, sensor in the system is used to build a proxy for the first sensor.
The two sensors can be measuring different phenomena or properties.
A virtual sensor is in some sense a model, but it is built strictly from data,
vice an explicate system model that is based on equations that describe the system.
Virtual sensors have two major uses:
1. Monitor unreliable sensors.
2. Replace expensive sensors with cheaper measuring devices.
As an example
(apologies for the poor figure quality -- we'll fix it soon),
An example of a virtual sensor.
Steam flow rate in a power plant is used to
predict the water level in the pressurizer.
The technique is simple and relies on the fact that in a nonlinear system all the variables are coupled.
The embedding theorem the guarantees that there is a relationship of the form
VB(n) = fB(y(n))
where VB(n) are the measurements from the sensor to be modeled and y(n) is the state space representation of
the data collected at the reference sensor.
The function fB is not known in general. It is derived as needed by expanding:
for the neighborhood around y(n).
Procedurally, baseline data for both sensors is collected over some representative period of time
(yes, you must collect data from the second sensor).
Using software for the analysis of chaotic data,
the dimension (degrees of freedom), dE, is determined.
When a new observation for the reference sensor is acquired, the formulas above are then used to make an estimate
of VB(n).
Return to chaotic.com Home Page.
http://www.chaotic.com/virtual.html
© Copyright 2006, 2007, 2008 chaotic.com®
Last modified 1/6/08