中文摘要：

将改进过的人工蜂群算法用于训练人工神经网络，对紫外光谱重叠严重的邻氯酚、对氯酚以及2,4-二氯酚的混合溶液进行同时测定。在260—290nm的范围内，使用经过正交设计的25组混合溶液的吸光度作为神经网络的训练集。当网络的误差平方和收敛到0．1时，输入另5组混合溶液的吸光度预测其浓度。对照实验表明，使用新算法训练的神经网络在回收率和收敛速度上与使用BP算法、粒子群算法以及标准人工蜂群算法相比均有较大提高。

英文摘要：

Artificial neural network trained by improved artificial bee colony algorithm is used for simultaneous quantitative analysis of seriously overlapped ultraviolet spectrum of a o-chlorophenol, p-chlorophenol and 2, 4-diehlorophenol. In the test, the artificial neural net- work are trained with the absorbancy of 25 groups mixed solution designed by orthogonal method which are detected in 260-290 nm. The other 5 groups mixed solution are input to forecast their chroma when the sum of square error of network converge to 0.1. Control experi- ments show that new algorithm outperform than BP, particle swarm and standard artificial bee colony algorithm on recovery and conver- gence rate.