X-Ray Fluorescence Spectroscopy Combined With Discriminant Analysis to Identify Imported Iron Ore Origin and Brand: Application Development
LIU Shu1, ZHANG Bo1,2, MIN Hong1, AN Ya-rui2*, ZHU Zhi-xiu1, LI Chen1*
1. Technical Center for Industrial Product and Raw Material Inspection and Testing,Shanghai Customs,Shanghai 200135,China
2. Department of Chemistry,College of Science,University of Shanghai for Science and Technology,Shanghai 200093,China
Abstract:Iron ore is an important raw material for the iron and steel industry. China is an iron ore import-demand country and the world’s largest iron ore consumer. The main goal of the customs’ inspection of imported iron ore is to prevent the risk of safety, health, environmental protection, fraud and other aspects of imported iron ore. The compliance verification of the origin and brand of imported iron ore can quickly screen the phenomena of adulteration, adulteration, and inferior charging, which support the risk management of imported iron ore and ensure trade facilitation. This article expands the application based on previous research. The research objects are 422 imported iron ore samples from 5 countries. In this paper, the accuracy of the non-standard sample analysis method of wavelength dispersive X-ray fluorescence spectrum is investigated. For the elements not detected in the measurement process, the detection limit was chosen to replace the missing values. For the outliers in the measurement process, F-test based on residual variance is used to eliminate the outliers. Each of the Pilbara Blend Lumps, Newman Blend Lumps, and Newman Blend Fines has one F statistic calculated from one set of data is greater than the F-test critical value (a=0.01), so these three sets of data are eliminated. The contents of Fe, O, Si, Ca, Al, Mn, Ti, Mg, P, Na, Cr, K, Sr, S, Zn, V, Cu, Ba, Ni, Mo, and Pb are selected by the stepwise discriminant method as the characteristic variable of the original identification model, and a four-dimensional Fisher discriminant model is established to identify the origin of the iron ore. The contents of Fe, O, Si, Ca, Al, Mn, Ti, Mg, P, Na, Cr, K, Sr, S, Zr, Zn, V, Cu, Ba, Cl, Ni, Mo, and Pb are selected by the stepwise discrimination method as the feature variables of the brand recognition model, and a 20-dimensional Fisher discriminant model is established to realize the recognition of 21 brand iron ores. The contribution of characteristic elements to the classification and recognition model is investigated, and the element characteristics of misidentified brand iron ore are analyzed. On this basis, the paper summarizes the whole data processing flow of the discrimination analysis model of the origin and brand of imported iron ore.