Research On Identification Of Silk/Polyester In Different Proportions Based On Electronic Nose

Silk is a kind of natural animal protein fiber,known as “human body second skin” and “fiber queen”.The characteristics of silk fiber have endowed quilt with excellent performance,such as health care,air permeability,and the promotion of sleep.However,the simple production process and low doorsill lead to the phenomenon of fake and poor quality commodity,which endangers the healthy development of the industry.Electronic nose is a kind of intelligent detection instrument which simulates the human olfactory system and has been widely used in various fields in recent years.Recognition mechanism of the electronic nose technology is that sensor array in 10 sensors is sensitive to different sensing material.Ultimately,The E-nose is through different pattern recognition methods to distinguish different gases.This topic is based on the production the status of silk quilt and the problems of silk fiber content,and obtains the data set of different proportions of silk/polyester samples by E-nose.The pattern recognition method was used to distinguish different proportions of silk/polyester yarns mixed samples by extracting different eigenvalues.By means of multilayer perceptron neural network and discriminant analysis method to predict the silk content in the unknown mixed samples which gives a new method of the identification of silk content fillings.The main results are as follows:(1)The samples of Zhejiang silk and Jiangsu silk province were analyzed.The cluster analysis result showed that the four types of silk in different producing areas and different time periods,the 20 samples divided into two categories,and the classification accuracy was 95%.Therefore,the different regions and different time periods silk sample can be classified into the same class.(2)The outside experimental parameters affecting the response of the electronic nose sensor were studied using the analysis of variance analysis,minimum significance analysis and relative standard deviation the optimal test parameters.Finally,the space of the empty space was 800 ml,and the sample quality was 3g,and Overhead generation time was 45min-60 min.(3)The principal component analysis and discriminant analysis were performed on the electronic nose response signal values of different moments.The steady state value is a more effective information characteristic sample set,which is beneficial to the performance of the later pattern recognition classification.Therefore,in order to improve the performance of the classifier,the steady-state value was analyzed and the mode file was established.(4)In the detection of different silkworm samples,the contribution rate of the sensor S2,S6,S7,S8 and S9 was found by the analysis of sensor response signal value and sensor load analysis.The first two principal components contribution rate was 97.33%,which was about 9% higher than the previous contribution rate.The factor loading matrix was orthogonal with the variance analysis method.Sensors of S7,S2,S9 had a higher loadings on the first principal component.Thus,the first common factor can be explained as containing sulfur compounds and nitrogenous compounds,while the sensors of S8,S6 had a higher loading in the second principal component,and the main substances are alcohols and aromatic compounds.(5)In this paper,the data set of steady-state value characteristic parameters and 14 parameter data sets were analyzed respectively.In 15 s,45s and 75 s,the recognition accuracy of the samples of 14 selected parameters was 89.27%,and the accuracy of the verification set was 88.57%.A linear discriminant analysis was performed on the steady-state data of the sensor after optimization,and the detection rate of the training set was 91.43%,and the detection rate of the verification set was 90%.(6)The correct rate of the training set of the multilayer sensing neural network was 91.07%,and the correct detection rate of test set was 90.78%.It is shown that multilayer perceptive neural network can be used as an effective method to detect the silk content of silk quilt.The research ideas of this topic have some guiding effect on the identification of silk in the future.It is helpful to improve the speed and precision of silk products detection and reduce the production of human error,which can help enterprises control the quality of silk products.