Virtual sensing

Virtual sensing techniques,^[1] also called soft sensing,^[2] proxy sensing, inferential sensing, or surrogate sensing, are used to provide feasible and economical alternatives to costly or impractical physical measurement instrument. A virtual sensing system uses information available from other measurements and process parameters to calculate an estimate of the quantity of interest.

While a variety of virtual sensing techniques are available, the vast majority of these can be classified in two major categories:

• Analytical techniques
• Empirical techniques

Analytical virtual sensing

Analytical techniques base the calculation of the measurement estimate on approximations of the physical laws that govern the relationship of the quantity of interest with other available measurements and parameters. For well-understood processes these approximations can be very accurate (e.g. using mass & energy balance equations) while for other processes precise physical models do not exist and the used approximations can be quite crude. Analytical virtual sensing is often implemented through data validation and reconciliation methods.

Empirical virtual sensing

Empirical techniques base the calculations of the measurement estimate on available historical measurement data of the same quantity, and on its correlation with other available measurements and parameters. The historical data of the un-measured quantity can be deriving either from actual measurement campaigns with temporarily installed sensor systems, or from detailed estimations with complex physical models that are computationally too expensive to run on-line. Empirical VS is therefore based on function approximation and regression techniques that can be implemented using a variety of statistical or machine learning modelling methods, such as, to name a few:

• Linear regression
• Weighted least squares (WLS) regression
• Kernel regression
• Regression trees
• Support vector regression
• Neural network regression

References