Department of Chemistry, Yazd University, Yazd, Iran.
Spectrochim Acta A Mol Biomol Spectrosc. 2011 May;78(5):1380-5. doi: 10.1016/j.saa.2011.01.014. Epub 2011 Feb 26.
The acidity constants of 4-(2-thiazolylazo)-resorcinol (TAR) were determined by the principal component-wavelet neural network (WNN). Biprotic acid mass balance equations, the distribution functions and the corresponding spectral profiles which were generated by a Gaussian model, have been considered to simulate all required absorbance-pH data. The simulated absorption-pH data matrix was used as training set whereas the TAR absorption-pH data was used as the test set of WNN model. The obtained acidity constants were in good agreement with the reported values of acidity constants in the literature and with those calculated by DATAN software. Artificial neural network (ANN) model has been also employed in this study and the results of WNN were compared with those obtained by ANN. It was found that WNN gives faster convergence and slightly better accuracy.
用主成分-小波神经网络（WNN）法测定了 4-（2-噻唑偶氮）间苯二酚（TAR）的酸度常数。采用双质子酸质量平衡方程、高斯模型生成的分布函数及相应的光谱轮廓，模拟了所有所需的吸光度-pH 数据。模拟的吸光度-pH 数据矩阵作为 WNN 模型的训练集，而 TAR 的吸光度-pH 数据作为模型的测试集。得到的酸度常数与文献中报道的酸度常数以及 DATAN 软件计算的酸度常数吻合较好。本研究还采用了人工神经网络（ANN）模型，将 WNN 的结果与 ANN 的结果进行了比较。结果表明，WNN 具有更快的收敛速度和稍高的准确性。
