A Data Mining based Intelligent System for Worsted Process Decision

Abstract

Textile production is a very complex industrial process, whose planning still depends on experts' knowledge and experience. With traditional techniques, a great many process parameters have to be repeatedly computed and the optimization of process parameters is also getting more and more difficult. However the proliferation of a huge mass of data from real production has been creating many new opportunities for those working in textile science, engineering and business. The field of data mining (DM) and knowledge discovery from database (KDD) has emerged as a new discipline in engineering and computer science. This paper investigates data mining methods from the industrial database, and presents a novel DM-based intelligent model (DMIM) for worsted process decisions through an integral application of case-based reasoning (CBR) and artificial neural network (ANN) techniques. First, from the rich existing process database, CBR is able to retrieve and recommend a similar process case as a process template; then, by means of modification on these parameters in the existing cases, ANN model is used to predict the yarn quality and make the best process decision. The basic concept and system modeling are presented in this paper. An applied case with DMIM is also given to demonstrate that the best process decision can be made and important process parameters such as those for raw materials can be optimized.