Abstract: Combination of wavelet transform with function of data compression and neural network is firstly applied to predict quantitatively the concentration of Pb²⁺ and other ions in oscillographic chronopotentiometric determination(OCPD). Compared with RPNN and WNN, combination of the wavelet transform and neural network(CWTNN)for OCPD has higher prediction accuracy and less convergence epoch. This can be explained from two aspects. Firstly, the network operation rate has been greatly enhanced because the optimal detail signal obtained after the wavelet transform not only is the characteristic information in original signal, but also has less number of data points than that of original signal. Secondly, the higher prediction accuracy can be obtained because detail signal used as network input contains lower noise. As a result, the potential application of CWTNN for OCPD and other electroanalytical methods will be very excellent.

Key words: oscillographic analysis, oscillographic chronopotentiometry, chemometrics, wavelet transform, neural network

摘要： 首次将具有数据压缩功能的小波变换与神经网络相结合用于Pb²⁺等金属离子的示波测定。与反弹传播神经网络、小波神经网络相比，本方法具有更高的预测精度和更少的收敛迭代次数。这一方面是因为使用经小波压缩后的信号作为神经网络的输入，压缩后的信号不仅提取了原信号中的特征信息，而且使网络输入的数据点数大幅度下降，大大提高了网络的运算速度。另一方面，由于选用了较高次分解所得的高频部分作为网络输入，从而即使在原始信号中含有较高的嗓音时也能获得较高的预测准确度。因此，将具有压缩功能的小波变换与神经网络相结合的方法必将得到广泛的应用。

关键词: 示波分析, 示波计时电位法, 化学计量学, 小波变换, 神经网络