Design Of Microplastics Separation Device And Study On Microplastics Recognition Model Based On Roman Spectroscopy

Microplastics refer to plastic particles fibers,and fragments with a particle size of less than 5 mm.As a new type of pollutants,microplastics have been paid close attention to by more and more people at home and abroad.Microplastics which have a small particle size and strong chemical stability are difficult to degrade,and they can exist in the environment for a long time.Due to the enrichment and migration of microplastics for typical pollutants,they will threaten both food safety and human health.In recent years,related reports on microplastic pollution have increased rapidly,and relevant research works are also being carried out.How to extract microplastics from environmental samples and determine the types of microplasics in the separates is the key to the research.In view of this situation,this paper designed a simple and efficient separation device that can separate microplastics from sediments,and then established a recognition model to identify the microplastics.The main research contents are as follows:1.Design of separating device.According to the principle of floating separation,a separating device was designed,and it has several advantages,such as simple operation and high separation efficiency.Based on relevant experiments,it was found that using NaI as a separating medium could effectively separate microplastics from sediments.2.Raman spectrometer was used to collecting and pretreating the spectral information of 9 kinds of microplastics.Through investigating the influence of integral time,scanning range and other parameters on spectral information,the optimal collection parameters were selected.The pre-processing of the collected spectral information mainly includes smoothing,baseline calibration,normalization and use principal component analysis to reduce the dimension of spectral data,so as to provide convenience for subsequent modeling.3.BP neural network model,support vector machine recognition model and Fisher discrimination were adopted to recognize microplastics.For BP neural network model,the optimal network parameters were as follows: number of input layer nodes:7,number of hidden layer nodes: 6,number of output layer nodes: 4,frequency of training: 500,learning rate: 0.2.The SVM model had the best effect when it adopted radial basis function and [0,1] normalized data processing method training model.For Fisher discrimination,a typical discrimination function was established.The model's stability was tested by self-validation and cross-validation of the correction set,and the discrimination rate of self-validation and cross-validation was 100% and 88.9%,respectively,which proved that Fisher discrimination analysis model was feasible for the identification and classification of microplastics.