Overview

When critical measurements are slow to reflect process changes or only lab analysis is available, then parameter estimation can often be used to improve the performance of control and monitoring applications.

The old saying, "You can't control what you can't measure," has haunted many Advanced Process Control projects. Being able to either accurately measure or predict an important process parameter is an integral part of closing the loop, but one that is often afforded little attention. There are many approaches available in building properties estimators for closed-loop control. Clearly understanding the final control objective of the property estimator and matching this objective with the best estimation technique is often the key to a successful control application.

The basic concept of inferred properties estimation involves using the measurable inputs of a process and their relationship to some property, such that this property is inferred or estimated from the values of the measurable inputs. The final use of this inferred property may take one of many forms, including:

• Estimation of an unmeasured physical property

• Validation of an online analyzer

• Generation of a continuous signal with feedback from a sampled analyzer

• Continuous estimation of a laboratory quality measurement

• A steady-state predictor to remove delay and lag from a process measurement.

Along with the many possible uses of the estimator, there are just as many techniques available to build the relationships and implement the estimator online. The proper selection of the estimation technique is an important step. It is often possible to...