论文部分内容阅读
针对传统化学物相分析对浸取剂高选择性要求及严重串相问题等缺点,以稳定的浸取常数为基础,建立了用于铅相态分析的数学模型,将聚类法与反向传播(BP)神经网络相结合,用于模拟地质样品中铅的相态分析。对隐层节点数、学习步长、学习样本、过拟合现象等进行了探讨,为分析结果的准确性提供了保证。对5个模拟样品进行了分析,相对误差小于±4%,相对标准偏差为0.99%~1.55%,结果优于传统化学物相分析结果。
Aiming at the disadvantages of the traditional chemical phase analysis, such as high selectivity to the leaching agent and serious problems of the tandem phase, a mathematical model for the analysis of the phase state of the lead was established based on the stable leaching constant. Propagation (BP) neural network is used to simulate the phase state analysis of lead in geological samples. The number of hidden nodes, learning steps, learning samples and over-fitting phenomena were discussed, which provided a guarantee for the accuracy of the results. The five simulated samples were analyzed. The relative errors were less than ± 4% and the relative standard deviations were 0.99% ~ 1.55%. The results were better than the traditional chemical phase analysis.