Data-based design of inferential sensors for petrochemical industry

Inferential (or soft) sensors are used in industry to infer the values of imprecisely and rarely measured completely unmeasured) variables from online (e.g., pressures, temperatures). The main challenge, akin classical model overfitting, designing an effective inferential sensor is selection a correct structure sensor. represented by number inputs sensor, which correspond their (simple) combinations. This work focused on design for product composition industrial distillation column two oil refinery units, Fluid Catalytic Cracking unit Vacuum Gasoil Hydrogenation unit. As first step, we use several well-known data pre-treatment (gross error detection) methods compare ability these approaches indicate systematic errors outliers available data. We then study effectiveness various taking into account complexity accuracy resulting model. analysis indicates that improvements achieved over current up 19%.