Research On The Technology Of Automotive Plastics Identification Based On Principal Component Analysis Of Near-infrared Reflectance Spectroscopy

With rapid growth in automobile population in China, The consumption of automotive plastics is also increasing constantly. The recycling of plastic for vehicles scrapped becomes more and more important. The dismantled plastic parts of vehicles scrapped should be classified for recycling, however, the issue of identification remains one of the key technical issues in respect of recycling of plastic for vehicles scrapped. Commonly used methods for site identification of plastic include apparent identification, density identification, refractive index identification, combustion identification, dissolving identification, static friction identification and other identification methods. It is difficult to identify automotive plastics accurately by using these methods. Although the Mid-infrared spectroscopy can be used to quickly identify the automotive plastic, it has higher requirements for the experimental environment, and cannot be used to quickly identify on the dismantling site.The scope of analysis by the near-infrared spectroscopy almost covers all organic compounds. The near-infrared spectroscopy has advantages such as rapid analytical speed, high sensitivity, good reproducibility and environmental adaptability, etc., but there are still problems such as failing to recognize dark plastic. Based on near-infrared spectroscopy technology,this study carry out in-depth research for recognition of scrap automotive plastic using near infrared spectroscopy technology and achieve the accurate and fast identification of scrap automotive plastic purposes.In the case of the issue the near-infrared spectrum failing to identify dark plastic, this study combine the long and short wavelength band near-infrared spectrometers, which not only use the advantages of short wavelength near-infrared spectrum such as low noise and high spectral quality, but also give full play to the advantages of the long wavelength band near-infrared spectrum such as rich information, high sensitivity, being help to improve the measuring accuracy of the component accounting for smaller proportion in the complex system, and have developed the near-infrared spectrum identification technology for the plastic for vehicles scrapped covering 900-2500 nm band. In order to resolve the influence of the surface state of the materials, color and other factors, this study has optimized the calibration model repeatedly and expanded the differences on the spectrum characteristics of the plastic with different surface state, thereby having improved the ability of the equipment to predict dark plastics for the scrapped vehicles and the accuracy of identificationFor the characteristic information about the near-infrared diffuse reflectance spectrum of a variety of the plastics for vehicles scraped is easy to overlap in the principal component space and the distinction is not strong, by improving features extraction algorithm of chemometric, effectively remove the interfering factors having nothing to do with the nature of samples such as the noise of the near-infrared diffuse reflectance spectrum of a variety of the plastics for vehicles scraped, band overlapping and baseline drift; extract characteristic information of the plastics for vehicles scraped; select the appropriate principal component space dimension, expand the differences between the characteristics information about the near-infrared spectrum of a variety of the plastics for vehicles scraped, and have respectively established long and short near infrared band calibration models for the plastics for vehicles scrapped, having improved the accuracy of detection.Finally, on the basis of the calibration model for the near-infrared spectrum of the plastic for vehicles scrapped, we use the Mahalanobis distance discriminant method in the cluster analysis to classify the unknown plastic for vehicles scrapped, thereby maximizing the homogeneity of the characteristic information about the samples of the plastic for vehicles scrapped in the same type and maximizing the heterogeneity of the characteristic information about the samples of the plastic for vehicles scrapped in different types, and then we can realize the purpose of accurate identification of the plastic for vehicles scraped.We develop the near-infrared spectrum modeling identification software by combining the major component analysis and Mahalanobis distance discriminant method, and the accuracy rate for identification of nine kinds of plastic for vehicles scrapped such as PP, PE, PVC, ABS, PS, POM, PMMA, PA and PC is up to 98%.This study reached the requirement of identification of scrapped automotive plastics on the dismantling site.