Prediction of wood pulp K with radial basis function neural network

Abstract

This paper presents a radial basis function (RBF) network for prediction of continuous wood pulp delignification factor. In pulp making process, the quality of pulp is measured by the K# which is related to the lignin content remaining in the pulp. Availability of an accurate K# during any time of digester operation is very critical in quality control and saving of million of dollars by reducing energy and raw material consumption. To assure quality control in the digester operation, K# is currently measured by human experts who analyze pulp samples in the plant's laboratory and then decide how to control process variables. Unfortunately, the long measurement time of K# coupled with process time delays has proven to be too costly economically and qualitatively. In our study, a radial basis function (RBF) neural network was implemented for prediction of K#. This network was trained with 350 and tested with 325 hours of real digester operation. Excellent results were achieved.