DEVELOPMENT OF INFERENTIAL PROCESS MODELS USING PLS

Inferential variables are often used in process industries in place of direct on-line measurement of controlled variables where direct measu rement is expensive, unreliable or adds significant delay. Simplified fundamental models are often not available for inferential control; th erefore, empirical models must be used. The procedures currently used for building empirical inferential models are based on standard statis tical methods and are generally limited to only a few preselected vari ables. This work investigates the use of a multivariate regression met hod, Partial Least Squares or Projection to Latent Structures (PLS). I t is shown that PLS provides a general method for building inferential models when one has data on a large number of process variables and w hen these variables are highly correlated with one another. By not ove rfitting the data PLS provides models with good predictive power, and through its very efficient handling of missing data, it provides infer ential models that are extremely robust to sensor failure. Since empir ical models are usually developed directly from process data, the natu re of the data set is extremely important. The data set must capture t ypical variation in all input variables and process disturbances. Furt hermore, the data collection must be designed according to the end use intended for the model. If the model is to be used in an inferential control scheme, then it is shown that open-loop process data cannot us ually be used. Rather, it is important that the data be collected unde r a feedback scheme that resembles the final scheme as closely as poss ible. Two case studies from distillation column control are used to de monstrate the general development of inferential models via PLS, and t o illustrate these points.