Data-driven Method For Controlling Yran Hairiness In Spinning Process

Spinning process is one of the important links in the entire spinning process.The hairiness produced by this process has a significant impact on the performance,quality and subsequent processing of the yarn,such as affecting the appearance of the fabric,easy pilling,and affecting the production rate and defects.Hairiness H value is one of the evaluation indexes of hairiness,and studying the hairiness H value is of great significance for exploring yarn hairiness and improving yarn quality.The generation characteristic of hairiness is the result of the superposition of the influence of each component of the spinning frame.It not only involves many influencing factors,but also has obvious periodic characteristics.This paper studies the periodic pattern recognition method of hairiness H value,and builds a prediction model of hairiness H value based on its periodic characteristics and long-term trend,and finally studies the quality traceability of hairiness origin.This paper mainly completed the following work:(1)In order to extract the periodic pattern from the periodic yarn hairiness H value data(that is,the change of the hairiness data on the cycle time or the length of the periodic data),a hairiness H value periodic pattern recognition based on dynamic time warping(DTW)is proposed Method,using local brute force search and pruning algorithm to optimize performance at the same time.The experimental results show that the actual cycle of the hairiness H value of the cotton spinning sample is 9.24 m;the theoretical cycle and the actual cycle of different types of yarns are different,and the average difference is 0.48 m;the cycle mode of the same type of yarn on different equipment can be used.abnormal detection.Compared with other identification methods,this method is simpler and more effective.(2)In order to improve the accuracy of online detection of quality indicators such as hairiness H value,this paper combines the short-term cycle and long-term trend characteristics of hairiness to construct an LSTM2 prediction model.First,use two sets of LSTM to predict the long-term trend and short-term cycle respectively,and then use the fully connected layer to synthesize the results of the two LSTMs,and finally get the model output.Through the comparison of experiments and statistical models and other neural network models,LSTM2 not only has the characteristics of continuous prediction,but also has superior prediction performance.The R2 score of LSTM2 is 0.7551,which is higher than all other models in the experiment.(3)In order to realize the data tracing of the causes of quality problems such as hairiness,a data tracing method based on Elasticsearch is proposed.This method aims at the backward manual identification system in the whole spinning process,and improves data traceability by digitizing the identification and standardizing storage.At the same time,the data traceability performance is improved by establishing a two-way traceability method and designing a reasonable traceability and fusion path.The results of comparative experiments show that this method has better storage stability;the traceability speed is 1.6 times and 1.2 times that of the My SQL and SQL Server implementation methods,respectively;the data format adopted is more flexible and scalable.This method has high application value in improving the quality traceability of hairiness and other quality evaluation indicators.This paper studies and determines the periodic pattern of yarn hairiness H value.The proposed periodic pattern recognition method makes an important contribution to the study of the mechanism of hairiness data characteristics and anomaly detection.The LSTM2 model established in this paper based on yarn hairiness characteristics is a combination of actual industrial scenes and data characteristics,and at the same time makes an important contribution to the development and progress of online hairiness index detection technology.Finally,this article traces the data on the cause of hairiness to locate the problem factors of unqualified hairiness,and looks for possible causes from the perspective of the front spinning process and data changes,which greatly improves the correction efficiency of the spinning mill.