Difficult-to-Measure Process Variable Estimation Based on Plant Data

Important process variables which give information about the final product quality cannot often be measured by a sensor, but their value is determined based on laboratory analysis. In order to enable a continuous monitoring of a process variable and an efficient process control, it is necessary to estimate this difficult-to-measure process variable. This paper deals with the appropriate methodology for building a suitable process model based on plant data, taken from the process database. Regression methods based on the input space projection into a latent subspace are proposed to build a model. Some properties of neural networks which make them a good basis for data based model building as well as for realization of the difficult-to-measure process variable estimator are pointed out. The properties of some proposed methods for process model building are demonstrated by modeling the crude oil distillation process based on the measuring data available.