Application Of Near Infrared Spectroscopy Analysis In Rapid Identification Of Common Plastics

Plastic is the most commonly used material in our daily life and industrial production.There are many kinds of plastic,different plastic has different characteristics,in order to meet the needs of different occasions.However,the massive use of plastic results in a lot of plastic waste.Plastic waste is different from ordinary domestic waste,and it can cause bad effects on the environment.The best way to deal with plastic waste is recycling.Because of the difference between different types of plastics,plastics generally cannot be used in combination.After the collection of plastic waste,the most critical step is to classify it so that it can be reused better.In this paper,we study the application of Near Infrared Spectroscopy analysis in rapid identification of common plastics.Near infrared spectroscopy(NIR)is a simple,fast and nondestructive technique for detecting sample information by measuring the transmissivity or reflectance of different wavelengths of the sample.This technique can be used to detect the plastics efficiently and make a series of qualitative or quantitative analysis with very high accuracy.A total of 347 plastic samples were collected,including 39 PA samples,39 PVC samples,40 ABS samples,30 PE samples,16 PPS samples,11 PPO samples,27 PS samples,38 PC samples,32 PET samples,51 PP samples,24 PBT samples,these 11 types of plastic would be used as research objects.Near infrared spectroscopy was used to collect samples near infrared spectrum,the acquisition range was 900-2500 nm,and finally,347 near-infrared reflectance spectra of 11 kinds of plastic samples were obtained.In order to eliminate the irrelevant interference and redundant information in the near infrared spectrum,and improve the signal-to-noise ratio of the spectrum,we use four different preprocessing methods to deal with these spectrum,including smoothing,derivative correction(first derivative and second derivative),multiple scattering correction and standard normal variate correction.On the basis of this,the principal component analysis was used to determine the best pretreatment method of plastic near infrared spectrum is to make the first derivative.The results of principal component analysis are used to establish the identification model and realize the effective identification of different kinds of plastics.Two different methods are adopted in this paper.The first method is to use pattern recognition,established the identification model based on BP neural network,probabilistic neural network and support vector machine,the prediction accuracy of unknown plastic samples reached 92%,91% and 94%,the results show that these three models can identify different kinds of plastics effectively.The second method is to establish a discriminating model based on the characteristic wavelength reflectivity ratio.Firstly,the four wavelengths of 1135 nm,1215nm,1670 nm and 2160 nm are characterized by the principal component analysis loading plot.After analysis,it is found that 11 kinds of plastics can be distinguished by comparing the reflectance ratio of 1215 nm and 1670 nm and the reflectance ratio of 1135 nm and 2160 nm.Based on this,the identification model was established,and the model was validated by using the prediction set plastic as an unknown sample.The results showed that the accuracy of identification was 98%,which proved the feasibility of the method.In summary,Using Near Infrared Spectroscopy analysis technique can effectively distinguish the 11 kinds of commonly used plastic,include PA,PVC,ABS,PE,PPS,PPO,PS,PC,PET,PP and PBT.The best method is to establish a model based on the reflectance ratio of the characteristic wavelength,with an accuracy of 98%.