Soft Sensor Design for Distillation Columns Using Wavelets and Gaussian Process Regression

Process variables whose measurements using the available equipment seem impossible or difficult are estimated inferentially by some computer programs called "soft sensors". In this paper, a novel soft sensor algorithm is designed to estimate the concentration of benzene, the bottom product of a distillation column. Wavelet denoising and Gaussian Process Regression (GPR) are utilized in the soft sensor design. Comparing to the other soft sensor designs, this model benefits from a high accuracy and less modeling complexity. The simulation results show promising robustness against the measurement noise.