Study on classification of aluminum plastic packaging tablets for drugs based on SOM-FDA using XRF spectroscopy(基于SOM-FDA利用XRF对药品铝塑包装片的分类)

Abstract

This paper presents establish a method for rapid classification of pharmaceutical aluminum plastic packaging sheets. Energy dispersive X-ray fluorescence spectrometer (XRF) was used to inspect aluminum plastic packaging samples from 47 different drugs. The importance of elements were sorted by maximum relevance minimum redundancy (MRMR) algorithm combined with a self-organizing map (SOM) network clustering, and K-nearest neighbor (KNN) was used to process sample data. The aluminum-plastic packaging sheets of pharmaceutical products can be differentiated based on the types and contents of elements contained in the samples. The clustering result of SOM neural network indicated 9 categories, and the accuracy of the KNN classifier was 97.87%. It shows that X-ray fluorescence spectroscopy is a simple, fast, non-destructive, and highly sensitive method for analyzing materials. The established classification model is scientifically accurate and can provide assistance for public security organs in large-scale screening, determining investigation directions, and shortening investigation time.(建立了一种对药品铝塑包装片进行快速分类的方法。利用能量色散型X射线荧光光谱（XRF）仪，对47种不同的药品铝塑包装片样品进行了检验，结合自组织映射（self organizing map，SOM）神经网络聚类，通过最大相关性最小冗余（maximum relevance minimum redundancy，MRMR）算法对元素重要性进行排序，并利用最近邻（K-nearest neighbor，KNN）分类器处理样品数据。依据样品中所含元素的种类及质量分数的不同，对药品铝塑包装片进行区分。SOM神经网络聚类的结果为9类，KNN分类器的准确率为97.87%。X射线荧光光谱法操作简便快速、无损检材、灵敏度高。建立的分类模型科学准确，可为公安机关大规模筛选、确定侦查方向、缩短侦查时间提供帮助。)

Keywords

• x-ray fluorescence spectroscopy(x射线荧光光谱法)
• pharmaceutical aluminum plastic packaging sheets(药品铝塑包装片)
• self organizing maps (som) neural network(自组织映射神经网络)
• k-nearest neighbor (knn)(最近邻分类器)
• classification(分类)