Inferential Models in Non-Linear Multivariable Control Applications

Abstract

On-line product quality measurements suffer from excessive time delays and sampling intervals, which limit their effectiveness as inputs for automatic control. Inferential models which are based on fast and continuously available temperature, pressure and flow measurements reduce the negative impact of the sample intervals and time delays with minimum compromise on accuracy. The resulting continuous and fast response keeps the product qualities on specifications and minimizes the quality giveaway. This chapter provides the development procedure of correlation-based inferential models for a wide variety of refinery processes, use of inferential models in both open and closed loop environments, laboratory update mechanism, operators acceptance and benefits of inferential models and their use as a controlled variable in non-linear multivariable control applications.