Natural Computation and Process Modeling in the High-performance Fiber Plant

Abstract

Modern fiber processes are capable of generating huge amounts of process data, especially with the increased use of in-line sensors, chemical analyzers and instrumentation. The driving force for snch measurements is the potential use in process control and product quality assessment. Process response databases are often extremely large in size, highly complex and non-selective. Novel adaptive learning techniques like artificial neural networks may provide a solution to assimilate on-line measured process-and property-data to relevant information, in order to maintain the process under control. In this century, advanced statistical techniques and new adaptive computation techniques will be tbe leading techniques in data analysis and will become an essential part of daily operations in tbe bigh-performance fiber plant.