系统聚类-BP神经网络吸光光度法同时测定地质样品中钼钨锡锑

SIMULTANEOUS ABSORPTIOPHOTOMETRIC DETERMINATION OF Mo W Sn AND Sb IN GEOLOGICAL SAMPLES BY THE APPLICATION OF HIERARCHICAL CLUSTERING METHOD COMBINED WITH BP-ARTIFICIAL NEURAL NETWOR

根据朗伯 比耳定律,建立了用于多组分分析的数学模型,将系统聚类法与人工神经网络相结合用于地质样品中钼钨锡锑的同时测定,其相对误差小于±12.5%。对神经网络隐层节点数的确定,样本及其个数的选择,以及如何防止过拟合现象的产生等问题进行了讨论。

Molybdenum, tungsten, tin and antimony in geological samples were determined simultaneously by the same chromogenic reaction system of SAPF-CTMAB. The interacting interferences including their similar chemical characteristics and serious overlapping of their absorption spectra, were resolved by the combining use of the hierarchical clustering method and the BP-ANN. A mathematical model was established on the basis of the (Beer′s) Law. Discussions on the determination of number of nodes of middle layer of BP-ANN, the selection of sample and its number, and the prevention of over-fitting phenomenon were given. Accuracy and precision of this method were tested, giving satisfactory results.