Modeling on the Worsted Fore-spinning Working Procedure
Based on the standardization of producing data gathering from the worsted mill, the relevance analysis and multivariate stepwise regression are performed; the important factors for the roving unevenness (R<,1>) and weight (R<,2>) are selected. According to the sequence of the important degree, the parameters are put into the BP sub-network group which is multi-input and one output, the four BP network models for R<,1> and R<,2> are established. The mean error percents (MEP) between the forecast value of the 10 groups of testing samples and the observed value are all below 4％. Using the 20 groups of data that do not participate for modeling to forecast the roving quality, the absolute average error percent (％) for R<,1> and R<,2> by the relevance analysis method are 2.63％, 2.95％ respectively. Meanwhile the correlation coefficients between them are 0.884 and 0.958 respectively. These results are superior then the stepwise regression method that selecting parameters for modeling, which they are 2.98％ and 3.21％ for the precision; 0.838 and 0.953 for the correlation coefficients between the forecast value and the observed value.
worsted fore-spinning working procedure BP neural network relevance analysis multivariate stepwise regression
LIU gui YU Weidong
Textile materials and technology laboratory, Donghua University, Shanghai, 201620, P. R. China Textile materials and technology laboratory, Donghua University, Shanghai, 201620, P. R. China Depar
国际会议
上海
英文
59-62
2007-10-15（万方平台首次上网日期，不代表论文的发表时间）