Self-Organizing Neural Nets: A New Approach to Quality in Textiles

Abstract

Until recently, textile quality assessment has been very difficult and mostly done by human experts. In the search for a more objective methodology, neural network techniques can be an excellent alternative. The first part of our paper presents a general introduction to neural network modeling techniques and focuses more extensively on the "topological mapping" model. The second part applies those techniques to textile quality assessment. Two problem cases are assessed—carpet wear and set marks. Both are objectively analyzed using self-organizing Kohonen neural networks. In both cases, the results are good and the system also indicates the objectivity of the human experts.