Deakin University
Browse

File(s) under permanent embargo

Mill specific prediction of worsted yarn performance

journal contribution
posted on 2006-01-01, 00:00 authored by Rafael Beltran, Lijing Wang, Xungai Wang
Different spinning mills use different raw materials, processing methodologies, and equipment, all of which influence the quality of the yarns produced. Because of many variables, there is a difficulty in developing a universal empirical/theoretical model. This work presents a multilayer perceptron algorithm (MLP) model for the purpose of building a mill specific worsted spinning performance prediction tool. Sixteen inputs are used to predict key yarn properties and spinning performance, including number of fibers in cross-section, unevenness (U%), thin places, neps, yarn tenacity, elongation at break, thick places, and spinning ends-down. Validation of the model on mill specific commercial data set shows that the general fit to the target values is good. Importantly, the performance of the MLP shows a certain degree of stability to different, random selections of independent test data. Subsequent comparison against the predicted outputs of Sirolan Yarnspec™ confirms the overall performance of the artificial neural network (ANN) method to be more accuratefor mill specific predictions.

History

Journal

Journal of the textile institute

Volume

97

Issue

1

Pagination

11 - 16

Publisher

Taylor & Francis

Location

Manchester, England

ISSN

0040-5000

eISSN

1754-2340

Language

eng

Publication classification

C1 Refereed article in a scholarly journal

Copyright notice

2006, Taylor & Francis