The Prediction Model Of Cotton Yarn Based On Neural Network

Due to the complicated structure of the yarn and the multi-process of yarn production,there is a complex nonlinear relationship among the fiber indicators,the parameters of the spinning and the yarn quality in the process of spinning.Using traditional mathematical statistical model is hard to describe clearly,some domestic textile enterprises even use the past experience to analysis and evaluation.This action vastly hinders the development of enterprises.With the development of science and technology,using artificial intelligent methods in the textile field is rising up gradually,the application of artificial intelligence can not only make up for the defects brought by the traditional methods,but also can take self-adaption adjustment to meet the needs of spinning process.It brings new reform to the concept of spinning enterprises.Combining the results of previous researches,This paper will list several classical algorithms in the artificial intelligence such as artificial neural network,particle swarm optimization algorithm,mind evolutionary algorithm,extreme learning machine to model cotton yarn quality prediction.The full text is divided into six chapters, the contents of each chapters show as follows:The first chapter is the introduction, this chapter mainly introduces research background and significance of the yarn quality prediction, research status is concerned too.This chapter also explains the reason to choose cotton fiber quality as the starting point,the feasibility of the yarn quality prediction based on artificial neural network technology,simply introduce the main research contents of this paper.The second chapter introduces the independent variable and dependent variable of the prediction model,the definition of the maturity, length, fineness, moisture regain, strength and impurities,and which yarn quality index is deeply influenced by these indicators. Also introduces three yarn quality indicators about the yarn strength, evenness, mixed impurities, and points out which are the main fiber factors affecting these indicators.The third chapter mainly introduces theory of algorithms, this chapter introduces the basic structure and characteristics of artificial neural network, structure of BP neural network and mathematical description,also lists the disadvantages of BP neural network,next introduces the principle of particle swarm optimization(pso) algorithm and mind evolutionary algorithm,and how to describe them with mathematical methods,the advantages of two algorithms,also points out the design steps to optimize BP network using these two algorithms. In the end, the design idea of extreme learning machine is proposed.The fourth chapter is the analysis of experimental results, in order to study the relationship between cotton property index and yarn quality index,this paper selects ten cotton property indexes includes fiber main length, maturity, breaking strength and evenness,short fiber content and fiber upper main length, metric count, total number of defects, the impurity rate, moisture regain as input factors; Yarn quality index, yarn strength, evenness CV%, nep impurity counts as the output, the use of various algorithms for modeling, and comparing the prediction results of various algorithms, concludes which algorithm has better prediction accuracy and stability.The fifth chapter uses grey correlation system combines MEA-BP and ELM algorithm, using the grey correlation degree to reduce ten input factors into six input factors, compare whether the precision accuracy of the MEA-BP model and ELM model has risen up after the dimension reduction.The sixth chapter is a summary and outlook of this article. The main work, contributions of this paper, and some existing problems were summarized,an expectation to the further study is given.