Abstract
This paper concerns a preliminary study of classifying the abnormal tension signals observed in the draw-texturing machines, which correspond to various types of abnormal events. The wavelet transform with inhomogeneous shift sampling is used to extract the feature vectors from the tension signals. Then, the self-organizing map (SOM) algorithm is performed on the feature vectors to carry out the unsupervised classification of the signals. The results agree with the classification by skilled engineers that is based on their experience.
