人工神经网络-伏安分析法同时测定邻、间、对二硝基苯

将反向传播算法的前馈神经网络用于导数脉冲伏安分析法同时测定邻、间、对二硝基苯。实验在盐酸－氯化钾－乙醇介质中进行，悬汞电极作为工作电极。通过对网络结构和参数的优化，加快了训练速度，提高了预测的准确度。用该法对邻、间、对二硝基苯混合物进行定量分析，预测的相对标准误差（ＳＥＰ）分别为４２６％，４９９％和４８６％。对人工神经网络（ＡＮＮ）和偏最小二乘法（ＰＬＳ）的结果进行的比较表明，ＡＮＮ法优于ＰＬＳ法。

Simultaneous Determination of o _, m _, and p _Dinitrobenzene in Mixture by Artificial Neural Network and Differential Pulse Voltammetry

The back propagation artificial neural network(BPANN) and differential pulse voltammetry (DPV)have been used in the simultaneous determination of o _, m _, p _dinitrobenzene in mixture. The experiments were carried out in HCl_KCl_EtOH medium by using a hanging mercury dropping electrode. The training speed and the predication accuracy can be enhanced by optimizing the network structure and parameters. The relative standard errors predicated for o _, m _,and p _dinitrobenzene were 4 26 %,4 99 % and 4 86 %, respectively. The results suggested that the ANN approach was better than the partial least square(PLS)method.