Chemical Engineering @ChemengFiles.Com

Any process that must be monitored requires sensors to find the variables of a process in real time. If required to monitor a large number of variables in the process, in the same proportion will require a large number of sensors for these variables. The sensors are costly, require maintenance, and can be calibrated become very sensitive, besides needing a hardware that connects with the monitoring unit. Would not it be better if we could dispense with a few sensors and still be able to monitor the whole process?

That is the goal of so-called "Virtual Sensors" or "Sensors Software", better known in the middle of the engineering control as "observers" or "Dear State." Observers are computer programs that from measurements obtained by a certain amount of actual sensors, make an estimate of the process variables that does not have its own sensor. The program reproduces the observed system giving information about the incentives (tickets) that receives the actual process. A second source of information useful for observers are monitoring the data it receives the results of the system (departures).

In a first stage of this "virtual monitoring" the real variables of the process and the respective variables that are being estimated may differ somewhat. This is very similar to when it solves a algebraic equation using a numerical method such as Newton-Raphson. In order to solve the equation is needed to make an assumption of what would be the result of the equation. Subsequently, the algorithm method of reporting the actual outcome. Thus, there is a need for an observer program which is supposed to be the value of the real variable in the process. It is not necessary that this assumption is true, a quality of the observer is that after a while is reporting the actual value. This quality is called "convergence".

A desirable quality of an observer is having a rapid convergence. Since it will be smaller as long as it is reporting a value that is not true. This speed of convergence can be "tunable" or adjustable. Not always a speed of convergence is desirable too high because the dynamics of the observer may fall into instability. This is very similar to when you're looking at a good television program that is of interest. If you change the channel so quickly that it is not enough time to assess the content per channel, never come to choose a specific channel.

How reliable is the estimate carried out by an observer? This will depend largely on the knowledge to take on the actual process, because based on this information, is designed algorithm estimate. However, in some cases is very accurate estimation is not required, since the results of a monitoring system can be fed to a robust control system only need an approximation of the true values of the process to keep it under control. Similarly, while estimation algorithms are developed are often evaluated on their performance by comparing its estimate regarding actual measurements. That to have a degree of reliability once they dispense actual sensor.

The virtual sensors have found a wide field of action in chemical engineering. Mainly in the operation of reactors where chemical commonly requires monitoring of various chemical species in the system. In some other processes such as the use of biochemical reactors observers is not only a matter of economics but also of feasibility, since it is virtually impossible to have sensors that can identify the different species in line microbiological involved in the reactive system.

The virtual sensors or observers have noted a revolution in engineering control. Their study over the coming years these technological developments in the field of process engineering likewise revolutionize the industry of chemical processes.

For more information:
en.wikipedia.org / wiki / Statistical_estimation
http://en.wikipedia.org/wiki/State_observer
en.wikipedia.org / wiki / Kalman_filter