YARN TENSION PATTERN RETRIEVAL SYSTEM BASED ON GAUSSIAN MAXIMUM LIKELIHOOD

摘要

The unusual yarn tension inspection plays an important role in yarn quality measurement. The patterns of unusual tension, which are captured and recorded by an on-line yarn tension monitor system, can ensure precise recognition of the unusual type of tension in order to solve the problem immediately. However, it is not easy for operators of twister machines to recognize the patterns of unusual tension without related training. The traditional on-line yarn tension monitor systems only detect unusual variation in tension, but cannot identify the patterns of unusual tension for operators, especially in the improved quality yarns. To assist the operators in the pattern recognition of unusual tension, in this paper, we propose an unusual yarn tension retrieval system based on Gaussian maximum likelihood classification algorithm. The proposed system uses four features to describe the tension patterns, and includes the following processes: pattern generation, feature calculation, similarity degree measurement, new class detection and pattern retrieval. Experimental results show that the proposed system can serve as an efficient and fast tool to identify unusual tension patterns.