A Study Of Near Infrared Spectrum Identification Of Waste Plastics

In recent years, with the rapid growth of petroleum industry, the production andconsumption of plastics have a mushrooming in the worldwide. And now it hasbecome a part and parcel of people’s daily life. With the mushrooming of plasticproducts’ consumption, the waste plastics are constant growth. It also brings seriousdamage to the environment and affects environment balance and people’s life. So therecovery and utilization of waste plastic become the urge demand of society whetherfrom the aspect of improving environment or the aspect of resource conservation.This paper mainly analyzed the recycling markets and the technique of wasteplastics and its development prospect. It also introduces the near infrared reflectancespectroscopy and the qualitative analysis method of NIR. On this basis, the nearinfrared spectrum of6kinds of waste plastics were gathered, and then useing theprincipal component analysis to extract the characteristic wavelengths which were asthe inputs of later models.Next, a Fisher discriminant model was set. This model used the characteristicwavelengths which were extracted through the principal component analysis as theinputs. Then a discriminant function was got. Testing the stability of the model bymeans of calibration sample set itself validation and cross validation. The rate ofaccuracy was96.5%and89.5%respectively. And there were3mistakes whenclassified the waste plastics.Third, artificial neural network was set to identify the6kinds of waste plastics’near infrared spectrum. This article studied the BP neural network and PNNrespectively, and made a contrast between the two kinds of method. In particular, theaccuracy of BP neural network’s training model is98.84%, but its’ accuracy offorecast sets is just only73.33%. The accuracy of PNN’s training model is100%, andits’ accuracy of forecast sets is up to97.67%relative to the BP’s. It basically meetsthe need.Finally, an online identification device was designed. And in order to combine thesoftware and hardware and also lay the first stone for further applications.