Application of Artificial Neural Network to the Identification of High Performance Liquid Chromatographic Peak Purity
应用人工神经网络鉴定高效液相色谱峰纯度

In this work, artificial neural network is first presented to identify the purity of high performance chromatographic peak. The network uses modified back propagation model with fast covergence. The basic theory of this model is given in detail. The network learns the knowledge from the different spectra of the front peak part. then predicts the chromatographic values of the tail, peak part and compares these values with the original ones. The chromatographic peak purity is estimated according to the criterion, namely percent error between the calculated values and the original ones. This criterion is defined as 10% based on the experimental results. The factors affecting the limit of detection, such as, solute concentration, chromatographic resolution and degree of similarity of the spectra are investigated. The results show that this method is mainly influeuced by the degree of similarity of spectra and the other two factors have small effects. The limit of detection of this method is 5%. The satisfactory results show that this new method has certain theoretical and pragmatic value.