Soft Computing Models For The Correlation Of Test Results Of Raw Silk Cohesion

China has the largest production of silk in the world.Raw silk is an important component of textile exports in China.During raw silk trades,raw silk properties influence the procedure and results of the trades directly.The raw silk cohesion has a close relationship with the raw silk production process and influences the processing and properties of silk products.At present,the Duplan cohesion machine is widely used to test the raw silk cohesion.However,the test result contains quite a few human factors.The raw silk cohesion automatic test system developed by Soochow University can reduce the human factors greatly.But whether the test results of these two methods have a good correlation is unclear.In this thesis,the correlation model of the test results of raw silk cohesion of the raw silk cohesion automatic test system and Duplan cohesion machine is established using the soft computing methods.This thesis consists of three parts:(1)The neural network based correlation models of the test results of raw silk cohesion are established using Back Propagation Neural Network,Generalized Regression Neural Network and Radical Basis Neural Network,respectively.Therefore,Radical Basis Neural Network is utilized to establish the neural network based correlation models of the test results of raw silk cohesion.The computation results of the neural network model show that the apparent diameter differences of raw silk when rubbing nine and ten times measured with the raw silk cohesion automatic test system have better correlations with the cohesion frequency measured with the Duplan cohesion machine under the data conditions of this thesis.Statistical test results indicate that there is no significant difference between the apparent diameter differences of raw silk between rubbing nine and ten times.(2)The fuzzy mathematics based correlation models of the test results of raw silk cohesion are established using the fuzzy comprehensive evaluation and fuzzy nearness methods.This model consists of the fuzzy comprehensive evaluation and fuzzy nearness.The weights of the correlation characteristic number of fuzzy comprehensive evaluation and the correlation characteristic number of fuzzy nearness are determined with a variation coefficient and dispersion maximization integrated method.The correlation characteristic number of fuzzy mathematics for test results of raw silk cohesion is then obtained.The computation results of the fuzzy mathematics model show that the apparent diameter differences of raw silk when rubbing seven and eight times measured with the raw silk cohesion automatic test system have better correlations with the cohesion frequency measured with the Duplan cohesion machine under the data conditions of this thesis.(3)The computation results of the correlation of test results of raw silk cohesion of the neural network and fuzzy mathematics model are fused.The soft computing based correlation model of the test results of raw silk cohesion is established.Effect of the weight difference on the final correlation characteristic number is analyzed.It is seen that the weight difference between the correlation characteristic number of neural network and correlation characteristic number of fuzzy mathematics does not cause the final correlation characteristic number similar to the orrelation characteristic number of neural network eventually.The ranking results of the final correlation characteristic number show that the apparent diameter differences of raw silk when rubbing nine and ten times have better correlations with the cohesion frequency.Statistical test results indicate that there is no significant difference between them.Analyzing the physical background of this subject,the optimum test index of the raw silk cohesion automatic test system which has a better correlation with the cohesion frequency measured with the Duplan cohesion machine is determined to be the apparent diameter differences of raw silk when rubbing nine times.The correlation between the test results of raw silk cohesion of the raw silk cohesion automatic test system and Duplan cohesion machine is studied using the neural network and fuzzy mathematics methods in this thesis.The optimum test index of the raw silk cohesion automatic test system which has a better correlation with the cohesion frequency measured with the Duplan cohesion machine is determined.The research results lay a foundation for the status establishment and extending application of the raw silk cohesion automatic test system and have practical significance in the objective test of the cohesion property of raw silk.