摘要
建立了人工神经网络解析自然伽马能谱的模型, 该模型选用隐层BP神经网络结构, 利用自然伽马能谱信息, 对铀 (U) 系和钍 (Th) 系各开两个能窗, 单能峰钾 40 (40 K) 开一个能窗, 另从低能(100 keV) 到高能(3 000 keV) 开一个很宽的能窗, 这6 个能窗的计数及其6 个相对误差作为网络结构输入共12 个节点; U, Th 和K的含量及相对误差作为输出共6 个节点; 隐层12 个节点。用自编解谱程序对训练样本集10 条已知含量的自然伽马谱进行训练后, 使用该模型解析了12 条待分析的样品谱, 得到较好的结果。
A model using ANN in interpreting natural gamma ray spectrometer is established The hidden layer ANN BP network structure is selected and information on natural gamma ray is utilized Twelve nodes (six energy windows and six relative errors for the windows) are inputted as network structure Two windows for uranium's and thorium's, respectively, one window for single energy peak value of k 40 and a very wide window for energy ranging from 100 keV to 3000 keV are made There are six nodes (content of U, T and K, and their relative errors) to output and twelve nodes of hidden layers After training ten natural gamma spectra in sample set with interpreting spectrum program,12 sample spectra analyzed are interpreted with fairly good result
出处
《江汉石油学院学报》
CSCD
北大核心
1999年第4期39-40, ,共2页
Journal of Jianghan Petroleum Institute
基金
中国石油天然气集团公司 "八五"储备项目
关键词
自然伽马能谱测井
能谱
神经网络
BP模型
natural gamma ray spectrometry logging
energy spectrum
nerve network