Artificial neural networks applied to potentiometric acid–base flow injection titrations

Artificial neural network (ANN) was applied for data treatment as a multivariate calibration tool in a potentiometric acid–base flow injection titration. A multilayer feed-forward ANN model, with Levenberg–Marquardt weight error correction was used for data modeling. The neural network parameter architecture was optimized to establish a relationship between the titration profile and the acid concentration. Citric and malic acids in synthetic sample mixtures and in orange juices were analyzed and the performance of ANN was compared with that of partial least squares (PLS) regression.