Virtual Analyzers of Mat Quality Based on Slurry Sensors

Abstract

Mats are often composites of fibers and filler particulates. They are formed by draining a dilute aqueous slurry on a filter medium called forming fabric, consolidated and dried. In general. formation of non-woven fibrous mats containing fibers and inorganic filler particulates is strongly influenced by the drainage characteristics of the slurry. These characteristics depend on chemical additives as well as physical properties of the slurry constituents for example size distribution of fibers and fillers. The qualities of resulting mats. such as wet strength and water load, arc important control parameters from the viewpoint of princess runnability and efficiency. Proper control of mat quality parameters allows one to enhance productivity through scrap reduction. Using on-line slurry analyzer data in combination with the processing variables of the forming section. we can obtain estimates of trial quality parameters in real-time through applications of neural networks. We use neural networks to identify non-parametric functional approximation, for predictive process modeling. This paper deals with real-time prediction of wet mat quality using on-line monitoring of slurry characteristics. This is significant as it allows one to obtain a product property that is normally available from destructive laboratory tests only.