Soft Sensor Design Using PLS and Neural Networks (NN): Comparison and Industrial Application
Arun K. Tangirala, CPC Group, U of A
CPC Group Seminar, February 10, 1999 (CME 572, 12:00 p.m.)


Abstract

The Melt Flow Index (MFI) provides a convenient measure to characterize polymer processibility. In the polymer manufacturing operations, it also serves as an indicator for product quality control. Mitsubishi Chemical Corporation makes several polymer grades in their Mizushima plant using the high pressure, high temperature autoclave technology. In this work, we summarize the results of a university-industry collaboration relating to the development of a soft sensor for the MFI of the polymers produced. Partial Least Squares (PLS) and Neural Nets (NN) are employed to obtain an estimate of the infrequently available MFI measurements from the routinely available process measurements. A comparative study of both these techniques in issues such as quality of predictions, the ease of modelling, etc. is presented.

This paper was presented at the AIChE '98 poster session.