Inferential Control

Many variables which characterize the quality of the final product in a chemical process are often difficult to measure in real-time, and hence cannot be directly used in a feedback configuration for process control. 'Inferential' control has come to cover of such problems from the case where the primary output is not measurable, to the case where the primary variable measurement is available at a lower frequency (and often time delayed) relative to the rapid secondary measurements. In general, inferential control algorithms cover both the construction of estimates of unmeasured outputs as well as estimates of unmeasured disturbances.

In regard the inferential control, “Nonlinear inferential control for process applications” has been published in Journal of Process Control and I found it very interesting.