Soft Sensor Design for Prediction of Product Quality Attributes

A. Giles, L. Wang, and C. Chessari (Australia)


Product Quality Measurement, Soft Sensor, Time Delay, Control System


Online product quality monitoring is an important issue for the Pet food industry. Sensors capable of measur ing product quality attributes such as Bulk Density and Moisture Content are inherently expensive. As a result, these measurements are often only available through off line analysis in a laboratory. This lack of available data measurement combined with the complicated and non linear nature of a food extruder makes on-line monitor ing and control of product quality attributes very diffi cult. This paper explores the design and validation of a soft sensor for predicting a product quality attribute in an industrial extrusion process. A continuous time transfer function model is built between the product quality at tribute and a set of on-line measured variables. A recur sive estimation algorithm is utilized to adapt the estimated models to changes in feed material and/or the plant op erating condition. Industrial experimental results demon strate accurate prediction of the product quality attribute.

Important Links:

Go Back