Abstract:

In order to establish a method to test the material evidence of plastic steel windows， 40 different brands， different series of the same brand， and different batches of the same brand and the same series of plastic steel windows were inspected by X-ray fluorescence spectrometer. The Niton XL3t GOLDD+ hand-held fluorescence spectrometer was used to determine the detection conditions： Ag as anode target； detection voltage： 50kV； detection current： 200μA； acquisition time： 70s. Under these conditions， samples can be detected according to the type and mass fraction of characteristic elements measured by different plastic steel window samples. In order to make the classification results scientific and accurate， the principal component analysis method was used to reduce the dimension of the experimental results， and four factors were designated and extracted. Then， the sample data were clustered by using the hierarchical clustering and K-means fast clustering methods. 40 samples were divided into 28 categories， and Fisher discriminant analysis was used to verify the scientific rationality of the classification results. Combined with spectrogram analysis and chemometrics， this method can classify samples quickly， accurately and non destructively， and provide help for the police grassroots to handle cases.

Key words: plastic steel window, physical evidence, X?ray fluorescence spectrometry, dimension reduction, cluster analysis