Establishment Of Identification And Classification Model Of Waste Plastics Based On Near Infrared Spectrum

Over the past few decades,the demand and production of various plastic products have been greatly increased with the great progress of science and technology and the rapid development of economy and society.However,a large amount of plastic solid waste are produced every year with the increasing consumption of plastic products.The traditional incineration and landfill treatment have caused serious pollution to the ecological environment,and also caused great waste of resources.Therefore,carrying out the research on recycling of waste plastics is the urgent need to achieve the sustainable development of economy and society.This paper mainly analyzes the present situation and development prospects of recycling waste plastics at home and abroad,the commonly used plastic recycling technology,and the research of identification of waste plastics using the spectrum.On this basis,this paper proposes the establishment of identification and classification model of ten kinds of waste plastics based on near infrared spectroscopy technology.Secondly,this paper discusses the working principle of the near infrared hyperspectral imaging technology and the feasibility of identifying the waste plastics.The near infrared spectral images of PE,PP,PS,PC,PA,PU,PET,PVC,POM and ABS plastics are acquired using near-infrared imaging spectrometer.The spectral data in the region of interest are extracted from the spectral images after the correction of black and white.Thirdly,the calculation principle and the derivation process of principal component analysis and partial least squares regression are discussed in detail in this paper.Taking PE,PP and PET these three kinds of the most common plastics for example,the characteristic wavelengths are selected using these two methods,and the merits and faults of these two methods are verified using the results of the Fisher discriminant analysis.The results show that eleven characteristic wavelengths are selected based on the principal component analysis,and the accuracy of self test and cross-validation are both 100 % during the process of discriminant analysis.Ten characteristic wavelengths are selected based on the partial least squares regression,and the accuracy of self test and cross-validation are 100% and 99% during the process of discriminant analysis.Therefore,the principal component analysis is selected as the method of selecting characteristic wavelengths.Fourthly,the calculation principle and the derivation process of the distance discriminant analysis and support vector machine sequence minimal optimization algorithm are described in detail in this paper.Identification and classification models are established using spectral data corresponding to the eleven characteristic wavelengths of PE,PP and PET these three kinds of plastics which are selected by the principal component analysis,and the classification accuracy is 100% and 97.14% respectively.Then,the classification accuracy of these two models are verified by thirty prediction set samples.The results show that classification accuracy rate of the models which are established by these two methods are both 100%.Therefore,it is preferred to establish the identification and classification model that is used for predicting unknown sample based on the distance discriminant analysis method.Finally,multi-level identification and classification model is established of the ten kinds of waste plastics using the principal component analysis and distance discriminant analysis,and the classification accuracy of this model is verified by prediction set samples.The result shows that there are four samples entifying incorrect during the one hundred and ten prediction set samples.The classification accuracy rate of this model reachs 96.36%,and it meets the requirements of identification and classification accuracy rate basically.Ultimately,a set of identification and classification process of unknown plastic sample is designed in the end of this paper.